Citation
Altruistic punishment and compensation

Material Information

Title:
Altruistic punishment and compensation results from a third-party justice game in Papua New Guinea
Creator:
Foreman, Rachel D
Place of Publication:
Denver, CO
Publisher:
University of Colorado Denver
Publication Date:
Language:
English
Physical Description:
xiii, 126 leaves : ; 28 cm

Subjects

Subjects / Keywords:
Au (Papua New Guinean people) ( lcsh )
Altruism ( lcsh )
Altruism ( fast )
Au (Papua New Guinean people) ( fast )
Genre:
bibliography ( marcgt )
theses ( marcgt )
non-fiction ( marcgt )

Notes

Bibliography:
Includes bibliographical references (leaves 113-126).
Thesis:
Anthropology
General Note:
Department of Anthropology
Statement of Responsibility:
by Rachel D. Foreman.

Record Information

Source Institution:
|University of Colorado Denver
Holding Location:
Auraria Library
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
74818100 ( OCLC )
ocm74818100
Classification:
LD1193.L43 2006m F65 ( lcc )

Downloads

This item has the following downloads:


Full Text
ALTRUISTIC PUNISHMENT AND COMPENSATION:
RESULTS FROM A THIRD-PARTY JUSTICE GAME IN
PAPUA NEW GUINEA
by
Rachel D. Foreman
B.S., Sewanee: The University of the South, 2000
A thesis submitted to the
University of Colorado at Denver
in partial fulfillment
of the requirements for the degree of
Master of Arts
Anthropology


This thesis for the Master of Arts
degree by
Rachel D. Foreman
has been approved
by


Foreman, Rachel D. (M.A., Anthropology)
Altruistic Punishment and Compensation: Results from a Third-party Justice Game
in Papua New Guinea
Thesis directed by Associate Professor David P. Tracer
ABSTRACT
The results of a third-party justice game performed among the Au of Papua New
Guinea are presented. Contrary to the predictions of canonical economic and
evolutionary theory, but consistent with results from previous experiments, players do
not always behave selfishly: the bimodal offers are 30% and 40%, and the third-party
altruistically acts (to punish, compensate, or do both) 34.7% of the time. Players
punish and compensate equally at 13.0% of the time each, suggesting that both
altruistic punishment and altruistic compensation are means of sustaining cooperation
among the Au. It is argued that this is due to inequality aversion compounded by the
cultural influences of 1) a structured system of reciprocal exchange that places a
premium on generosity; and 2) the justice system of Papua New Guinea, a mixture of
the customary restorative justice system and the Western retributive justice system,
imposed on Papua New Guineans by colonial administrations. Results of this study
have both theoretical and empirical implications, as they foster debate about the
theories for the evolution of altruistic cooperation, and highlight the need for future
testing of such theory, especially toward refining understanding of the influences of
gender, fairness, and justice on game behavior.
This abstract accurately represents the content of tty£ candidates thesis. I recommend
its publication.
Signed
David'P. Tracer
iii


DEDICATION
I dedicate this thesis to Sherie, proof that altruism exists.
I also dedicate this thesis to my parents, Dianne, Paul, and Katie, for your
unconditional support and love, now and ever; and to Joshua, for unfaltering
understanding, support, and encouragement during the cold-shower days and
complaint-filled nights of writing.


ACKNOWLEDGMENT
My thanks to the John D. and Katherine T. MacArthur Foundation for funding, and
to David P. Tracer for the opportunity to travel to Papua New Guinea as his research
assistant on the project. Without the collaboration of Dr. Tracer and his sharing of
data, this thesis would not exist. Tenkyu tru. Gratitude is also due to the Department
of Anthropology, University of Colorado at Denver, for additional funding and also
for support and guidance throughout my graduate studies. Special thanks to Connie
Turner for all of her generosity and help the past two years. Many thanks also to Dr.
John Brett and Dr. Christopher Beekman for all of their guidance and many helpful
suggestions during the writing process. Finally, thanks to the people of Anguganak,
Papua New Guinea for their kindness and generosity, and because of whom I stayed
well fed of sweet potatoes and did not succumb to my bout of dengue fever or
malaria, whichever 1 had.


CONTENTS
Figures.....................................................................x
Tables.....................................................................xi
Maps......................................................................xii
Photographs..............................................................xiii
Chapter
1. Introduction and Specific Aims..........................................1
1.1 Central Question........................................................1
1.2 Justification of the Site..............................................2
1.3 Specific Aims and Predictions..........................................4
1.4 Significance...........................................................6
1.5 Chapter Summaries......................................................7
2. Theoretical and Empirical Background....................................9
2.1 Introduction...........................................................9
2.2 Human and Non-human Altruism..........................................11
2.3 Theories for the Evolution of Altruism................................14
2.3.1 Kin Selection........................................................14
2.3.2 Reciprocal Altruism..................................................17
2.3.3 Indirect Reciprocity and Costly Signaling............................20
vi


2.3.4 Altruistic Punishment
23
2.4 Behavioral Economics....................................................30
2.4.1 A Description and History..............................................30
2.4.2 Homo oeconomicus......................................................32
2.4.3 Games: The Prisoners Dilemma, Dictator Game, and Ultimatum Game.. 35
2.4.4 Weaknesses of Game Theory.............................................43
2.5 Conclusion..............................................................44
3. Study Population.........................................................48
3.1 The Setting.............................................................48
3.1.1 Papua New Guinea.......................................................48
3.1.2 Anguganak, Sandaun Province...........................................50
3.2 The People and Culture................................................51
3.2.1 TheAu..................................................................51
3.2.2 Reciprocal Exchange...................................................53
3.2.3 Law and Justice.......................................................54
4. Justice..................................................................57
4.1 Types of Justice........................................................57
4.1.1 Retributive Justice....................................................57
4.1.2 Restorative Justice...................................................59
4.1.3 Distributive Justice and Fairness.....................................60
4.2 Justice and Game Theory.................................................62
vii


5. Methods
64
5.1 Sample Recruitment....................................................64
5.2 The Game.............................................................65
5.3 Data Collection......................................................67
5.4 Analysis.............................................................70
5.5 Methodological Issues................................................71
6. Results and Discussion................................................74
6.1 Quantitative Results..................................................74
6.1.1 Frequencies of Variables............................................74
6.1.2 Player A Offers.....................................................76
6.1.3 Sanctions and Compensations.........................................77
6.2 Qualitative Results..................................................82
6.2.1 Interviews and Vignettes............................................82
6.3 Discussion...........................................................83
7. Conclusion and Future Directions......................................93
7.1 Conclusion..........................................................93
7.1.1 On the Evolution of Altruistic Cooperation..........................94
7.2 Future Directions....................................................95
viii
7.3 Concluding Remarks
97


Appendix
A. Map and Photographs..................................................98
B. Script, Data Sheet (Pre-game Interview),
Post-game Interview and Vignettes..............................101
C. Human Subjects Review Board Approval................................112
Bibliography.............................................................113
ix


FIGURES
Figure
1. Education by Gender.......................................................71
2. Education by Gender and Village..........................................71
3. Distribution of Selected Means...........................................76
4. Vanilla and Cocoa by Village.............................................76
5. Distribution of Player A Offer (Kina) and Player C Action................78
6. Distribution of Player C Action for Offers 0%-50%........................80
7. Vignette Responses.......................................................83
x


TABLES
Table
1. Example of a Prisoners Dilemma Payoff Scheme.........................36
2. Role by Village and Gender Totals.....................................70
3. Role by Gender and Village............................................70
4. Frequencies of Variables..............................................74
5. ANOVA Comparison of Variable Means Between Villages...................75
6. Frequency of Player A Offer (Kina) at Brugap, Winaluk, and Anguganak.... 77
7. Frequency of Player A Offer (Kina) and Player C Action
at Brugap, Winaluk, and Anguganak...........................78
8. Player C Action by Offer..........................................79
9. Player C Decisions Overall and by Gender..........................80
10. Summary of Mean and Modal Offers in Ultimatum Games..............84
xi


MAPS
Map
A. 1. Papua New Guinea..........................................98
Xll


PHOTOGRAPHS
Photograph
A. 1. A traditional-style house............................................98
A.2. A more modem, woven-walled house atop stilts.........................98
A.3. An even more modem house with corrugated roof........................99
A.4. Mama Opa scraping sago..............................................99
A.5. Sago scraping with a bamboo hammer...................................99
A.6. Meini washing sago, using her hand-made bamboo processor.............99
A.7. A typical meal of sago jelly and greens..............................100
A. 8. Market day at the Station...........................................100
A.9. Community school................................................100
A. 10. Sarah grooming Janet................................................100
A. 11. The author conducting pre-game interviews..........................100
A. 12. The principal investigator, David Tracer, conducting the game
with a third-party participant...................................100
xiii


1. Introduction and Specific Aims
1.1 Central Question
Altruismdoing an act that benefits another individual at a cost to oneself
is anomalous under the singular premise of both canonical economic and
evolutionary theory. This premise predicts that individuals are self-interested
maximizers of material gains. Sub-theories from varied disciplines have been
advanced in the attempt to explain how altruistic cooperation could arise and be
maintained among a population of egoists, and ultimately become evolutionarily
stable by natural selection. These theories include inclusive fitness, reciprocal
altruism, indirect reciprocity, costly signaling, and altruistic punishment. A subfield
of economics, game theory, is devoted in part to the empirical testing of these
theories with the aim of understanding altruistic behavior. This paper describes the
results of a novel game used to test such theoretical predictions.
Building on experimental economic games previously performed by Tracer
(2003, 2004, Henrich et al. 2001), this thesis presents the results of a third-party
justice game performed among the Au of Papua New Guinea. In brief, a proposer
makes an offer to an inert and anonymous recipient of how to divide an initial
endowment. An anonymous third-party is given an endowment equal to 50% of the
initial endowment, and the opportunity to sanction the proposer, compensate the
recipient, or do both, all at a cost to himself. These costs are a 20% sacrifice to
punish, 20% sacrifice to compensate, and a 40% sacrifice to both punish and
compensate. Conversely, if the third-party chooses not to act he leaves the game with
the full endowment allotted to him. This research design will test the basic economic
1


and evolutionary prediction that players will behave selfishly, test the validity of the
theory of altruistic punishment as a norm-enforcer, and explore the influence of
culturespecifically local ideas about justiceon game behavior.
According to the selfishness axiom, conventional evolutionary and
economic theory predicts that in the third-party justice game, the proposer will offer
nothing, and that the third-party will never act on any proposal, despite how unfair
the offer is to the recipient. However, empirical results from diverse experimental
economic games played both among the Au and elsewhere show that in economic
games, individuals do not strictly behave as rational, selfish, utility maximizers.
Instead, they altruistically offer as much as 50% of their endowment with or without
the threat of punishment; individuals also altruistically punish in games (Bolton et al.
1998, Bowles and Gintis 2004, Boyd et al. 2002, Camerer 2003, Cameron 1995,
Diekmann 2004, Eckel and Grossman 1998, Fehr and Fischbacher 2004b, Fehr and
Gachter 2002, Fehr et al. 2002, Fowler et al. 2004, Henrich and Boyd 2001, Henrich
et al. 2001, Hoffman et al. 1996, Nowak et al. 2000, Riechmann 2002, Romp 1997,
Roth et al. 1991, Tracer 2003). Moreover, results greatly differ between
industrialized and small-scale societies, ostensibly due to cultural mores about
fairness and the degree of free-market integration (Henrich et al. 2001). Because
fairness and justice are inextricably linked, the research design described here also
aims to elucidate whether the predominant justice system plays a role in player
behavior by allowing the option of punishment, compensation, or both actions at a
cost to the third-party. These actions may act as proxies for retributive, restorative,
and distributive justice, respectively.
1.2 Justification of the Site
Several characteristics make the Au an ideal population for testing game
theory. First, results from behavioral economic studies performed in small-scale
societies (Henrich 2000, Henrich et al. 2001, Tracer 2003) greatly differ from those
2


in industrialized nations (Camerer 2003, Cameron 1995, Diekmann 2004, Hoffman
et al. 1998, Roth et al. 1991). The study of cooperation in smaller-scale settings
like among the forager-horticulturalist Aumay prove more tractable than studies
performed in large-scale groups with complex division of labor. Results from small-
scale societies may provide valuable information about the pre-conditions for
cooperation among large-scale groups and thereby help refine theory about the
evolution of altruism. Second, typical of Melanesia, the Au have a structured system
of reciprocal exchange that influences social, economic, and political interactions,
and bolsters social ties within and beyond the family or clan. The enculturation of
reciprocal norms may produce interesting (or anomalous) experimental results that
may corroborate or detract from previous theoretical assertions about social
preferences either based on cultural rules of fairness, or on inequality aversion.
Indeed, Tracer (2003) reports that in economic games previously performed among
the Au, participants are not only very generous, sometimes offering greater than 50%
of the original endowment, they also altruistically punish with high frequencythat
is, at a cost to themselves, they punish those who make unfair offers. Further study
will help ascertain to what degree this behavior is due to the entrenchment of
reciprocal norms. Third, previous game theory experimentation in small-societies
and among the Au themselves (Henrich et al. 2001, Tracer 2003, 2004) provide both
a foundational guide for the research design reported here and a body of empirical
results for comparison. Fourth, game behavior had been said to be largely based on a
local sense of fairness (Henrich et al. 2001), and justice is implicit in ideas of
fairness. Therefore the justice system of a particular society may inform game
behavior. In addition to a Western retributive system, Papua New Guinea also
utilizes, at least customarily if not formally, a restorative justice system, namely in
the emphasis of victim compensation. In such a context, game experimentation that
gives the third-party player the option to punish the proposer and to compensate the
recipient, both of whom are anonymous, will help 1) determine how ingrained either
3


type of justice system is, and 2) support the idea that altruistic rewards may be just as
important as altruistic punishment for the maintenance of cooperation.
1.3 Specific Aims and Predictions
The following hypotheses make predictions about proposer (I.) and third-
party behavior (II.). These hypotheses flow from the selfishness axiom of both
economic and evolutionary theories.
I. The proposer will never make a non-zero offer.
If individuals are selfish, as economic and evolutionary theory predicts, offers in the
third-party justice game should not exceed 0%, especially under conditions of
anonymity.
The frame of the third-party justice game should likewise not induce non-zero
offers. The threat of punishment has been the alleged cause of non-zero offers in
ultimatum games (See Chapter 2). In that game, a proposer makes an offer to a
recipient of how to divide an endowment. The recipient may accept the offer, or he
may reject it so that neither player gets any of the endowment. Thus, in the
ultimatum game the threat of punishment to the proposer is direct, ostensibly driving
proposers to offer non-zero amounts to ensure they leave the game with a positive
payoff. In the third-party justice game however, the threat of punishment is indirect;
that is, it comes from a third-party, whose endowment is given independent of
proposerrecipient action. Because the recipient is inert, and because the threat of
punishment is indirect, offers should never exceed the purely selfish, 0%.
Cultural norms could, however, drive offers up from the predicted non-zero
mark. Among the Au, these influences include (probably a combination of) the
ingrained system of reciprocal exchange typical of Melanesian society, which places
a premium on generosity; the lack of anonymous interaction in daily life, which
would cause participants to believe that their actions in the game might one day
4


induce detrimental returns should their identity be found out; local ideas about
fairness; or even the customary justice system that values victim compensation
among other punitive measures.
II. Third-party players will never act because it is costly to do so.
According to economic and evolutionary theories, the third-party should never
punish or compensate because of the 20% cost to do so; that is, it would be an
altruistic act. The third-party should especially not choose to both punish and
compensate because of the 40% cost to do so.
However, empirical evidence suggests that there are exceptions to these
assumptions, and that in economic games, individuals sometimes punish other
players at a cost to themselves. Such altruistic punishment has previously been
rationalized on grounds of fairness; that is, a player responds negatively and punishes
a coplayer for what he sees as unfair (here, selfish) behavior (Rabin 1993, Fehr and
Gachter 1998). While negative reaction toward unfairness may explain games in
which proposer-action directly impacts punisher-utility, in this game the third-party
receives his endowment regardless of the proposers (fair or unfair) behavior.
Punishment has also been rationalized on grounds of indirect utility
maximization; that is, decreasing the utility of the proposer (and probably the player
with the most utility) increases self-utility relative to other players (Tracer 2003).
Though both explanations flout standard economic theory, relative utility
maximization is congruent with evolutionary theory.
Compensation cannot be rationalized on grounds of economic or evolutionary
theory. However, like non-selfish offers, compensation may be explained by cultural
proclivities that emphasize reciprocity and victim compensation. Punishment and
compensation are equally costly, at a sacrifice of 20% of the third-party endowment.
However, the customary criminal justice system emphasizes victim compensation at
5


least as part of punition. Unfair offers approximate petty crime more than they do
violent crime, which might warrant more stringent punishment. Therefore third-party
compensation could be explained, at least in part, by the impact of cultural norms on
generosity and justice. However, the concurrent act of both punishing and
compensating comes at an enormous costa 40% sacrifice of their endowmentso
that this behavior should never be observed.
1.4 Significance
By adding to a growing body of empirical evidence for the existence of
altruism, this thesis will foster debate about the standard predictions of canonical
economic and evolutionary theory. Experimental results presented here and
elsewhere show that players (and by extension, individuals) do not always behave
selfishly. These results suggest there is a need for revision of theoretical
generalizations, or at least refinement of theory to help explain how altruism can
become stable under expectations of selfishness. Reform of theory will have
implications in a range of disciplines including psychology, anthropology, sociology,
economics, and political science (Diekmann 2004).
The importance of theoretical refinement is not limited to academia, but also
bears implications in economic policy and constitution design (Fehr and Gachter
2002, Gowdy and Seidl 2004). Standard (Western) economic models based on the
selfish, rational, individualistic actor lead to a micro-foundations approach to
economic and social policy. This approach assumes perfect competition in
capitalistic, impersonal markets that can and will reach equilibrium. In reality, the
social nature of humans allows us to learn and base behavior on culture as well as
feed-back from other individuals. Acknowledgment of this social nature produces a
much less parsimonious yet more accurate model of economic reality, a model with
the potential for multiple equilibria. This acknowledgment necessitates a wholesale
change of assumptions about consumption. Informed policy based on such changes
6


may ameliorate growing global inequities produced from the forcing of neo-liberal
economic ideals on developing nations.
1.5 Chapter Summaries
Chapter 2 provides an overview of theories invoked to describe the evolution
of altruistic cooperation and the use of experimental economic games to test such
theory. Following an introduction (Section 2.1), the characteristics of altruism are
discussed (Section 2.2), including the question of whether or not it exists among non-
human primates. Section 2.3 details the four main theories for the evolution of
altruism, starting with Hamiltons theory of inclusive fitness (Section 2.3.1) and
Trivers theory of reciprocal altruism (Section 2.3.2). These theories rely on a strict
interpretation of evolution by individual selection and do well to explain how
altruism could arise among kin and among dyads with repeated interactions,
respectively. The similar theories of indirect reciprocity and costly signaling (Section
2.3.3) extend the above theories to describe the proliferation of altruism among
groups larger than two individuals. Section 2.3.4 addresses the theory of strong
reciprocity, or altruistic punishment, to describe the sustaining of cooperation in
anonymous interactions; proponents of altruistic punishment use the controversial
idea of group selection to explain its inception and evolutionary proliferation.
Section 2.4 describes the importance of economic game theory as a burgeoning
avenue of social science research into social preferences, beginning with a
background and history in Section 2.4.1 and a presentation of assumptions in Section
2.4.2. Section 2.4.3 critically describes and presents empirical evidence of the
prisoners dilemma, dictator, and ultimatum games. Section 2.4.4 briefly enumerates
the weaknesses of game theory, and is followed by a theoretical and empirical
conclusion in Section 2.5.
Chapter 3 presents a brief overview of the study population, the Au of Papua
New Guinea, providing description of the research setting (Section 3.1) at the
7


national (Section 3.1.1) and local levels (Section 3.1.2). Section 3.2 is a very brief
ethnography of the subject population (Section 3.2.1), followed by a description of
the reciprocal exchange system that pervades Au society as it does in most
Melanesian societies (Section 3.2.2). Section 3.2.3 concludes the chapter with an
exploration of law and justice in Papua New Guinea, emphasizing compensatory
measures as a manifestation of restorative justice.
Chapter 4 provides a more thorough examination of the three main types of
justice (Section 4.1), retributive (Section 4.1.1), restorative (4.1.2), and distributive
justice (4.1.3). The idea that justice and the underlying issue of fairness influence
game behavior is explored in the final section (4.2), providing a transition into the
current research design, which aims not only to explore the altruistic (or selfish)
behavior of participants, but also elucidate whether a particular form of justice
predominates in game behavior among the Au.
Chapter 5 provides details of the research design and methodology. Section
5.1 describes sample recruitment, Section 5.2 explicitly describes the third-party
punishment game, and Section 5.3 presents the mechanism of data collection. A
summary of the sample population by gender is presented for each of the three
villages where the game experiment was performed. The chapter ends with a
description of data analysis (Section 5.4) and a brief enumeration of methodological
issues (Section 5.5).
Chapter 6 consists of the empirical results followed by discussion.
Quantitative results make up Section 6.1, specifically including a presentation of the
frequencies of variables (6.1.1), proposer offers (6.1.2), and third-party actions
(6.1.3). Qualitative results are summarized in Sections 6.2, with the results of
interviews and responses to vignettes (Section 6.2.1). Discussion commences in
Section 6.3. The final chapter features a conclusion (Section 7.1), final summary of
the theories for the evolution of cooperation (7.1.1), suggested future directions (7.2),
and concluding remarks (7.3).
8


2. Theoretical and Empirical Background
2.1 Introduction
Pure altruisma costly act performed on behalf of a genetically unrelated
individualposes an evolutionary problem under the assumption that natural
selection predisposes all organismsand arguably the genes they carryto be
selfishly concerned with fitness and reproductive success (Axelrod 1984, Darlington
1972, 1978, Fisher 1992, Hamilton 1964). Altruistic cooperation is only a stable
strategy in a population entirely composed of cooperators. Though altruism could
arise as a mutation, the proliferation of altruism among a population of egoists is a
biological conundrum if altruistic behavior decreases the fitness of altruists while
increasing the fitness of beneficiaries (Boyd and Richerson 2005). Theories of
inclusive fitness (Hamilton 1964, 1972) and reciprocal altruism (Trivers 1971)
attempt to explain how altruism could proliferate in animals including bees, ants, and
vampire bats, but they fall short of explaining human cooperation in large groups of
unrelated individuals (Henrich and Boyd 2001). In fact empirical evidence from
experimental economics predicts that natural selection does not favor cooperation for
groups larger than ten persons without a high degree of relatedness (Henrich and
Boyd 2001). Other theories, like those of indirect reciprocity (Fishman 2003,
Mohtashemi and Mui 2003, Nowak and Sigmund 1998), costly signaling (Austen-
Smith and Banks 2002, Gintis et al. 2001, McAndrew 2002, Zahavi 1975, 1977), and
strong reciprocity (or altruistic punishment) (Bowles and Gintis 2002a, Boyd et al.
2002, Fehr and Gachter 2002, Fehr et al. 2003, Fehr and Henrich 2003, Fowler et al.
2004, Gintis 2000a, Gintis et al. 2003) attempt to take up where inclusive fitness and
9


reciprocal altruism leave off. Ancillary to the debate of the evolution of altruism are
questions about the universality of human altruism, or to what degree altruistic
behavior takes its influence from social, cultural, economic, and political contexts
that produce social norms like fairness.
Anecdotal and ethnographic evidence of altruistic human cooperation without
obvious external sanctions exists in many forms, from food sharing of hunted game
to warfare (Fehr and Gachter, 2002) to giving blood and charitable donations (Boyd
and Richerson, 2005). While some acts may be rationalized by the potential for
future reciprocity, in others, a public good may confer benefit on an individual who
contributes nothing at all. Furthermore, beneficiaries may belong to large anonymous
groups with little or distant (genetic or otherwise) relation to benefactors.
Benefactors may also act among those with whom they will never interact again
(one-shot encounters) so that prestige or reputation gains are minimal or non-existent
(Bowles and Gintis 2004, Boyd and Richerson 2005, Fehr and Gachter 2002). In
addition to anecdotal accounts, further evidence of human cooperation comes from
controlled economic experiments used to test game theories (Berg et al. 1995, Bolton
et al. n.d., Bolton et al. 1998, Cameron 1995, Clark and Sefton 2001, Diekmann
2004, Eckel and Grossman 1998, Fehr and Fischbacher 2004b, Fehr et al. 2002,
Henrich 2000, Henrich et al. 2004, Hoffman et al. 1994, Hoffman et al. 1996,
Riechmann 2002, Roth et al. 1991, Telser 1995, Tracer 2003, 2004, Wedekind and
Milinski 1996).
Altruism is a particularly fascinating and arguably universal aspect of human
behavior, thereby making it the subject of study in a diverse range of fields from
philosophy (Kitcher 1993) to justice, biology, economics, and anthropology. Unlike
other universal human activities that have obvious biological functions (e.g., eating,
sex), altruism has yet to be fully reconciled in terms of Darwinian natural selection
which should disallow its spread and stabilization. This opens the door for the
alternative explanations, proposed by sociobiologists (Fisher 1992, Wilson 1983) and
10


selfish-gene theorists (Dawkins 1976), of group and multi-level selection (Alexander
2000). Before proceeding to theories for the evolution of altruistic cooperation,
issues in defining altruism will be discussed.
2.2 Human and Non-human Altruism
A number of messy definitions fall under the heading of altruism. Most
generally, any act costly to ones self and beneficial to another is altruistic. But this
would include such basic and instinctual behavior like the care of offspring (Fisher
1992). Others definitions eliminate nepotistic acts of kindness or care for
descendants and relatives from the repertoire of altruistic endeavors, to define
altruism as a costly act on behalf of a genetically unrelated individual. A further
issue is whether to include reciprocity under the heading of altruism. If an altruistic
act is committed with the expectation of return then perhaps the contingent exchange
is better defined as reciprocal mutualism. However, Darlington (1978) argues that
any altruistic act is potentially reciprocal: a beneficiary may decide at a later date to
return a favor. This further complicates the dividing line between altruism and
reciprocity necessitating one further distinction on grounds of intent: an altruistic act
is one that is made without expectation of return.
Pure altruism is very rare or arguably non-existent among nonhuman animals,
though as mentioned before, various forms of cooperation are often lumped together
under altruistic headings including kin-related cooperation and mutualism. Axelrod
(1984) argues that though selfishness is the primeval, evolutionarily stable strategy1,
cooperation may evolve when individuals interact repeatedly and frequently. This is
most often enacted by related individuals, but may also arise among higher
Whereas the Nash equilibrium describes the best strategy in a given environment in economics, the
equivalent in behavioral anthropology is the Evolutionarily Stable Strategy (ESS), or the schema that
cannot be undermined in a given ecological niche unless something in the environment changes
(Maynard-Smith 1982).
11


organisms with sufficient cognitive ability to both recognize previous interactive
partners and to remember results of previous interactions so that the appropriate
response may be chosen. While Axelrod argues that mutualism is a primitive
example of altruism, others (Brosnan and de Waal 2002) contend that, because
mutualism confers spontaneous benefit on the actors, it is a category all its own,
separate from altruistic and reciprocal behavior typical of gregarious animals like
humans and other non-human primates (Silk 2004, Terborgh and Janson 1986),
among whom group living and constant, repeated social interactions set the stage for
cooperation.
There are many examples of cooperation from the non-human primate world.
These acts may or may not be altruistic, depending on the relatedness of group
members. Chimpanzees stage collective border patrols and seemingly organized
attacks resembling warfare on neighboring groups (Alexander 1979, Silk 2004);
vervet monkeys sound alarm calls when predators encroach and may come to the aid
of allies and kin when in danger (Alexander 1979, Seyfarth et al. 1984); captive and
wild cebus monkeys, capuchins, and chimpanzees share food (Brosnan and de Waal
2002, Conroy 2005, McGrew 1998, Silk 2004); macaque females defend unrelated
juveniles from harassment by other group members (Chapais et al. 2001); langurs
and howlers carry other females infants (Silk 2004); and most primates spend 10-
20% of their waking hours grooming group members. Primates also sometimes
engage in polyspecific cooperative behavior; for example, squirrel monkeys form
alliances with capuchins, and both species seek out associations with tamarins
(Terborgh and Janson 1986). Though inconclusive, it seems that some of the same
advantages to intraspecific group living may apply to groups of different species
(Reed and Bidner 2004), namely predator avoidance and increased knowledge about,
access to, and defensibility of food resources (Fleagle 1999, Terborgh and Janson
1986).
12


Among group-living species, Axelrod (1984) proposes the following potential
keys to cooperative tendencies: 1) complex memory, 2) information processing that
uses past effects to inform present action, 3) ability to predict the probability of
future interactions (with the same individual), and 4) enhanced ability to distinguish
between different individuals and their associated behaviors. The latter may in fact
be the most important attribute, allowing individuals to favor some and punish
others, depending on previous behavior, and therefore protect oneself from repeated
harm from defectors. These qualities are naturally difficult to assess and are thus
controversial among nonhuman primates.
The above primate social activities may be considered to be altruistic to
different degrees. With the exception of polyspecific cooperation, it seems that most
of these activities demonstrate cooperative behavior between genetically-related
individuals, so that these acts are not examples of pure altruism. Moreover, Brosnan
and de Waal (2002) assert that low-cost, opportunistic reciprocity should not be
categorized with high-cost, altruistic reciprocity. In other words, it is much less risky
to sometimes participate in low-cost allo-grooming than to risk ones life in the
protection of conspecifics from predators. The above primate activities might be
better categorized as cooperation based on high rates of association or mutual
tolerance, not on contingent exchange. This distinguishes primate behavior from
human cooperation that is generally based on score-keeping, or remembering
previous interactions (Brosnan and de Waal 2002, Silk 2004).
Humans not only cooperate with kin and those with whom they have repeated
interactions, but with strangers as well. Both seem to violate the selfishness axiom of
evolutionary and economic theory because though individuals can receive benefit
from mutual cooperation with other individuals, they can benefit even further by
exploiting the cooperative actions of others (Axelrod 1984). Defection in one-shot,
random encounters makes perfect evolutionary sense, if heritable will create a
population of defectors, and is an evolutionarily stable strategy. But defection
13


becomes detrimental when interactions are repeated and previous defection, if
remembered, can prompt punitive measures. If a cheater should meet again those he
has exploited, the cheated may punish the cheater.
In summary, there is a contentious continuum of altruism (Humphrey 1997),
with pure altruism at one endespecially that in one-shot, anonymous encounters
and kin-related care (e.g., parent-offspring care) at the other. Mutualism, reciprocity,
and cooperation lie somewhere in between, depending on the degree of relatedness of
actors, delay in receiving returns, and cost of the altruistic act.
A growing body of empirical evidence from experimental economics shows
that non-selfish behavioral preferences are not anomalous (Axelrod and Hamilton
1981, Camerer 2003, Cameron 1995, Clark and Sefiton 2001, Diekmann 2004, Eckel
and Grossman 1998, Fehr and Gachter 2000a, 2000b, 2002, Henrich 2000, Henrich
et al. 2001, Nowak et al. 2000, Roth et al. 1991, Tracer 2003). Altruistic behavior is
not confined to individuals, but may confer benefit to a group; thus it is sometimes
called prosocial behavior (Henrich 2001,2004, Gowdy and Seidl 2004). In recent
years economists, dissatisfied with inaccurate predictions from neoclassical
economic theory, have turned to anthropologists and evolutionary biologists for
theoretical explanations of observed behavior. Specifically, these economists seek
answers as to how cooperation is maintained among large groups of unrelated
individuals and in one-shot encounters (Henrich et al. 2004). The following is an
exploration of the various theories for the evolution of altruistic cooperation.
2.3 Theories for the Evolution of Altruism
2.3.1 Kin Selection
In 1964, Hamilton proposed a genetical mathematical model for the
evolution of social behavior, specifically behavior difficult to reconcile under the
principle of classical natural selection. His model follows the same general
mechanism of natural selection but is based on a refinement of the concept of fitness,
14


to encompass not just the fitness of an individual but that of an individual plus all of
his genetically related kin.
The unit of Darwinian natural selection is the individual, where fitness is
measured by the relative number of that individuals offspring that survive to
reproduce. Dawkins (1976) argues that findings in genetics have moved focus from
the individual to the heritable gene as unit of selection. It is gene preservation and
proliferation that is at the core of individual survival. This tweaking of the definition
of fitness bolsters the assumptions at the core of Hamiltons theory of kin selection.
When the unit of selection is not the individual but his genes, fitness and
reproductive success take on new nuances because an individual shares common
genes with his kin (Fisher 1992). Reproductive success then, becomes the additive
reproductive viability of both vertical (descendent) and collateral (non-offspring)
relatives (Alexander 1979, Alexander and Hamilton 1981, Daly and Wilson 1978,
Fisher 1992, Hamilton 1975). For the positive selection of a gene, an increase in
individual fitness is not enough if it comes at the expense of related individuals who
may carry replicas of the same gene (Hamilton 1962). For example, siblings share
50% (on average) of their genes so that protecting and helping a sibling protects 50%
of ones own genes2 (Daly and Wilson 1978, Dawkins 1976, Fisher 1992).
Conversely, a gene disadvantageous to an individual may undergo positive selection
if that gene confers advantage on relatives. Therefore, if an individual who has the
genes for altruism helps a relative, this gene is shared in the beneficiary relative, and
the beneficiary survives and reproduces, the altruistic gene will also be replicated
(Fisher 1992).
Inclusive fitness is a quantity incorporating the overall number of genes of
related individuals, so that a positive net increase in this quantity will null the
negative fitness cost to the altruistic individual (Alexander 1979, Chapais 2001,
2 In other words, it makes as much sense to feed ones siblings as it does to feed ones children (Daly
and Wilson 1978).
15


Chapais et al. 2001, Darlington 1978, Fehr and Gachter 2002, Hamilton 1964, 1975,
Rodman 1999, Silk 2004, Strier 2000). If altruism is the bestowal of reproductive
benefit (b) upon a beneficiary at some cost (c) to the benefactor (Chapais 2001,
Chapais et al. 2001, Daly and Wilson 1978, Strier 2000), the altruistic act may be
selectively advantageous if the beneficiary and benefactor carry an allele in common.
The probability (r) of carrying that allele is the coefficient of relatedness of the two
individuals3. For the allele to be positively selected, the cost of the beneficial act
must be less than the benefit multiplied by the probability of relatedness (c that the fitness of the allele is increased (Matessi and Karlin 1984). In other words,
the net cost-benefit of kin altruism depends upon three variables: 1) the genetic
relatedness of the two parties, 2) the ability of the beneficiary to convert benefits into
reproductive success, and 3) the cost to the benefactor measured relative to
alternative choices (Alexander 1979). The selective advantage endowed upon the
altruistic individual declines with relatedness, as it is less likely that the individuals
share any particular gene (Daly and Wilson 1978).
This idea of inclusive fitnessindividual fitness plus that of genetically-
related members carrying the same alleleshelps to explain evolutionary paradoxes
that seem at first to belie the principle of natural selection, such as sterile worker
bees, the phenomenon that Darwin himself identified as a near-fatal threat to his
theory (Daly and Wilson 1978, Darwin 1859, Fehr and Gachter 2002, Silk 2004). It
may also explain predator alarm calls in animals, food sharing, and grooming. While
all organisms are selfish individualists striving for survival through reproduction, kin
selection would have gregarious animals also behave as altruists to maximize the
survival (through reproduction) of genes shared with kin (Alexander 1979). The
3The coefficient of relatedness is unity for clones (twins), one-half for sibs, one-fourth for half-sibs,
one-eighth for cousins, et cetera, and zero for those whose relationship is negligible (Hamilton 1964,
1975). (These are possible, not absolute proportions, because the exact gene proportion is only known
in the case of identical twins.) Thus one should not sacrifice his own life unless to save more than two
brothers, four half-brothers, or eight cousins.
16


sterile worker bee cannot reproduce, but altruistically helps its fertile relative,
thereby making it appear that it is the family that is the unit of selection, when it is
really the gene, as shared by family members.
Inclusive fitness-driven behavior is phenotypically altruistic, but
genotypically selfish (Alexander 1979) where inclusive fitness may be maximized
through the favoring of more closely related individuals. This is different from the
controversial theory of group selection. Though altruistic acts among related
individuals could be termed group selection of the genes those individuals carry, it is
not selection for individuals, because those individuals are not wholly identical.
Hamiltons theory of inclusive fitness is a feasible model for the proliferation
of altruistic cooperation in small groups of genetically-related individuals (Boyd and
Richerson 2005). However, some argue that Hamiltons formula for kin selection is
far too simplisticmathematically as well as in the assumption that altruism could
be heritable via a single locus alleleand that it exaggerates the probable
effectiveness of kin selection (Darlington 1978, Matessi and Karlin 1984). A further
problem is that the theory seems to imply that individuals are endowed with a great
deal of rationality in choosing to help relatives, presumably with the (unconscious,
yet) ulterior motive of bestowing kind acts on those from whom future (genetic)
returns are most likely. In addition, inclusive fitness fails to explain altruistic
cooperation when group size increases or the relatedness of group members
decreases. Furthermore, humans continue to cooperate even in one-shot encounters
with strangers (Boyd and Richerson 2005, Henrich et al. 2004), and when
cooperation transcends the boundary between species (Fisher 1992).
2.3.2 Reciprocal Altruism
After inclusive fitness, the second solution to prosociality is the idea of
reciprocal altruism. Championed by Trivers (1971), reciprocal altruism is the
manifestation of an unconscious heritable trait that prescribes cooperation in dyadic,
17


long-term, repeated interactions (Axelrod 1984, Boyd and Richerson 2005, Bradley
1999, Fehr and Gachter 2002, Fisher 1992, Henrich et al. 2004, Silk 2004). In
contingent exchanges, two or more individuals, regardless of kinship, profit more
than the cost of individual altruism (Darlington 1978, Diekmann 2004, Fishman
2003). Returns may arrive at any time after the initial act (Fishman 2003).
Reciprocity and altruism differ in that 1) reciprocity is conditional fairness, as
opposed to the unconditional generosity of altruism; 2) reciprocity is not categorical
(i.e., all or none) but is dimensional, so that it is possible to reciprocate to a degree;
3) reciprocity is an obligation evoked by previous behavior, but returns may be of a
different currency than that in the original exchange (i.e., tit-for-tat = eye-for-an-ear
[or something else] or tat-for-tat = eye-for-an-eye); 4) the reciprocal norm does not
only apply to benevolent action; punishment may be the return; 5) reciprocity may be
driven by egotistic motivation, to get something in return (from material goods to
heavenly honor, which may explain seemingly altruistic charitable donations)
(Gouldner 1960). Trivers (1971) quintessential example is that of cleaner-fish, or the
small fish that trades cleaning of the mouth, teeth, and gills of a larger fish in
exchange for a meal (and not being devoured). Extending this model of contingent
exchange to humans, Trivers used game theoretic methods to argue that the
reciprocal strategy is willing cooperation in the first interaction followed by
conditional, continued cooperation based on previous partner action; this strategy is
also known as tit-for-tat (TFT) (Axelrod 1984, Bartholdi et al. 1986)4. In reciprocal
altruism, cooperation is responsive and based on compatibility more than on kinship
(Clark and Sefton 2001). Moreover, if there is a gene for reciprocal altruism, survival
4 In computer tournaments, TFT never defects first, but always defects immediately after an opponent
defects, and always cooperates immediately after an opponent cooperates (Bartholdi et al. 1986).
18


will be enhanced by natural selection acting on mutual benefits5. This is because
cheaters6 will be selected against if the act of cheating has future costs (i.e., being
shunned from future cooperation) that outweigh the immediate benefit of cheating
(Trivers 1971).
Dispersal strategies to proscribe inbreeding require many animals to live in
groups with unrelated individuals. The theory of reciprocal altruism posits that
unrelated individuals may cooperate with the expectation of returns at a later time
when interactions are repeated often enough for previous partners to be recognized
(Henrich et al. 2004, Strier 2000, Trivers 1971). Several pre-conditions enhance the
likelihood of success of reciprocal altruism, including a long lifetime (so that
reciprocal encounters may be frequently experienced and remembered), low
dispersal rates, high mutual interdependence (including extended parental care), and
the existence of stable social groups (Trivers 1971). While Trivers (1971) contends
that cleaning fish and alarm calls in birds are examples of reciprocity altruism, others
assert that animals are incapable of pure reciprocity among unrelated individuals
(Alexander 1979, Bradley 1999). Alleged reciprocal evidence from the non-human
primate world includes grooming among vervet monkeys, where grooming between
unrelated individuals increases the chances that an individual will respond to alarm
calls (Seyfarth and Cheney 1984); and the sharing of food among chimpanzees,
ostensibly based on past interactions (Fehr and Gachter 2002), with some evidence
suggesting that chimps do so in memory-based, partner-specific exchanges (Brosnan
5 Hamilton would argue, however, that by virtue of obtaining a return for an altruist act, the act is no
longer altruistic by definition (Humphrey 1997). It is simple reciprocity, or an eye for an eye.
Trivers would retort that in inclusive fitness, individuals do altruistic acts on behalf of genetically-
related individuals and therefore is not pure altruism by definition. Humphrey (1997) argues however,
that genetic returns of inclusive fitness are returns so that kin selection has an element of reciprocity
to it. The difference between the two then, is intent: the kin altruist most likely does not expect a
return, so that reciprocal altruism is not altruism because there is an expectation of return.
6 Following Trivers (1971), a cheater is an individual who fails to reciprocate, without referring to
intent or intending to imply that the individual is morally defective.
19


and de Waal 2002) so that sharing is more likely among chimps who have shared
food in the past. For example, if chimp A requests food from B, B will be more likely
to respond with aggression if chimp A has not shared with B in the past (de Waal
1991). Some argue that this behavior in chimps proves that reciprocation is buried
deep in human evolution (Fehr and Gachter 1998).
Though the costs and benefits incurred from acts of proposed reciprocal
altruism are difficult to measure (Brosnan and de Waal 2002, Chapais et al. 2001,
Strier 2000), theory predicts that cooperation among unrelated animals can arise
when they interact regularly and have the opportunity to adjust their (cooperative)
behavior according to previous experiences (Seyfarth and Cheney 1984). The
potential for these interactions occurs in various animal societies typified by group-
living, as shown in the examples above. However, it does seem that overwhelmingly,
most reciprocal encounters in the animal kingdom occur between genetically related
individuals, while among humans that is not the case.
Reciprocity may best be summed by the adage an eye for an eye,
encompassing both the concepts of positive and negative reciprocity (Fehr and
Gachter 1998). Equal returns for equal gains underlies both concepts, yet negative
reciprocity may be more aptly viewed as punishment or revenge while positive
reciprocity is driven by the desire to return kindness to those who have been
previously kind (Rabin 1993). Two problems remain: how can reciprocity arise
under conditions of natural selection if it does not begin among kin; and how are
cheaters recognized (Fisher 1992)? The second problem increases in importance as
group size increases, especially when accurate recognition of previous cooperators is
required (Henrich et al. 2004).
2.3.3 Indirect Reciprocity and Costly Signaling
Trivers (1971) defines two distinctions in reciprocity: 1) direct reciprocity
results when rewards come from the same individual who receives original benefit;
20


2) indirect reciprocity results when rewards come back from the cooperative social
group at large in the form of prestige or respect (Alexander 1979, Fishman 2003,
Mohtashemi and Mui 2003, Nowak and Sigmund 1998). In other words, though Mr.
X did not do a favor directly for me, he did so for Mr.Y and for Mrs. Z. I can then
infer Mr. X is likely to return a favor to me. While direct reciprocity requires
repeated dyadic interactions, indirect reciprocity may facilitate cooperation in larger
groups so long as information about individuals reputations may continually be
assessed. Some argue that while non-human primates may be capable of direct
reciprocity, indirect reciprocity is unique to humans (Alexander 1979).
Proceeding from the field of psychology, costly signaling is akin to indirect
reciprocity. In costly signaling, actions or objects are used to communicate some
non-obvious information about intention or personality to other players (Austen-
Smith and Banks 2002; Gintis et al. 2001; McAndrew 2002). Signaling has been
invoked to explain seemingly inefficient market activities like advertising and
factory strikes, and has been used in biology to explain morphological traits such as
peacock feathers and apparently maladaptive traits like heavy, large elk antlers.
Other examples include product warranties, college degrees from a prestigious
university (Camerer 2003), and large corporation or celebrity donations leaked to the
media Costly signaling is sometimes called competitive altruism because an
individual may perform acts of extreme philanthropy in times of plenty in order to
position themselves for access to resources in later times of need (McAndrew 2002).
In this way, costly signaling is a good strategy for inducing reciprocal altruism. But
costly signaling may apply to a range of social interactions like resource sharing, 7
7 For example, during the writing of this thesis, the nation of Kuwait issued a public costly signal in
the devastating aftermath of Hurricane Katrina (New York Times 2005). Dr. Anas Al-Rasheed,
Kuwaits Minister of Information, wrote a full-page ad in the New York Times to express his
countrys condolences to the victims of the hurricane and to pledge $500 million in aid for the
reconstruction of the Gulf Coast. Though he worded the pledge as a reciprocal offering in thanks for
U.S. help during the invasion of his country by Iraq and in the subsequent rebuilding of Kuwait, the
very public pledge doubled as a costly signal of Kuwaits political alliance and economic status.
21


defense, raiding, and the punishing of free-riders or norm-violators (Gintis et al.
2001).
The quality of the signaler is the genetic or phenotypic attribute that is
difficult for others to assess directly. Yet, it has effects on payoffs from social
interactions with the signaler: those who give benefits or give more benefits (those
who signal more intensely) are advertising their good qualities. This influences
future behavior of group members who might allocate payoffs to the signaler, for
example, choosing them as allies or mates, or choosing to defer to them in
competition. Costly signaling may also benefit the group by virtue of sharing social
information (McAndrew 2002).
Signals are an evolutionarily stable, sustaining force of cooperation if they
are less costly than the benefits accrued when others decode the signal and
reciprocate; and are too costly for cheaters to fake (Camerer 2003, Henrich et al.
2004). Four qualities of a costly signal are that 1) the behavior must be easily
observable, 2) must be costly to the signaler (in resources, energy, etc.), 3) must be
an honest indicator of some real trait (otherwise it is just cheap talk)(Austen-Smith
and Banks 2002), and 4) must provide some advantage to the signaler (McAndrew
2002). In order to be altruistic, costly signals must also lead to some advantage to the
observer.
Indirect reciprocity and costly signaling may maintain cooperation in groups
where cooperators build a reputation and do not necessarily obtain material returns
(Fehr and Gachter 2002). The theory of costly signaling asserts that cooperation
evolves because it is a signal of the group members quality as a mate, ally, or
competitor, and later facilitates the formation of advantageous alliances for signalers
(Gintis et al. 2001). Both indirect reciprocity and costly signaling add the important
element of addressing altruistic evolution among multiple members, not just between
dyads. Though indirect reciprocity and costly signaling go farther in the explanation
of cooperation among unrelated groups of people than do inclusive fitness and
22


reciprocal altruism, these theories do nothing to ameliorate the problem of one-shot
or anonymous cooperation where there is no chance for signaling and no chance to
observe previously cooperative behavior (Bowles and Gintis 2002a, Henrich et al.
2004); nor do they explain cooperation when groups are very large and assessment of
individual reputation is difficult or impossible (Diekmann 2004, Mohtashemi and
Mui 2003).
2.3.4 Altruistic Punishment
Cheaters who take advantage of a purely reciprocal system threaten to destroy
reciprocal cooperation. Punishing these cheaters, however, reinforces the altruistic
system, and has been called negative reciprocity (Fehr and Gachter 1998), strong
reciprocity (Gintis 2000a, Gintis et al. 2003), or altruistic punishment (Fehr and
Gachter 2002).
Strong reciprocators are cooperators who punish non-cooperators even if
punishing is costly, even in anonymous and one-shot circumstances, and even if the
cost of punishment is unlikely to be repaid (Bowles and Gintis 2002b, Gintis 2000a,
Gintis et al. 2003, Henrich et al. 2001). If this form of interaction is termed strong
reciprocity, then the reciprocal altruism championed by Trivers (1971) is weak
reciprocity (Gowdy and Seidl 2004). Some (Bowles and Gintis 2004, Boyd et al.
2004, Gintis 2000a) argue that strong reciprocators are more likely to survive when
their group faces extinction than either reciprocal altruists (who rely on repeated
interactions) or costly signalers (indirect reciprocators)8 as punishment could
maintain norms of cooperation better than these under duress. Self-interest with a
view to future success is at the core of both reciprocal altruism and inclusive fitness,
where self-interestedness allows natural selection to favor these mechanisms of
8 Arguably, altruistic punishment may confer a fitness advantage on social-groups with an above-
average number of punishers so that they may better survive group-threatening catastrophe, like war
and famine (Sigmund et al. 2002). Under these conditions, cooperation breaks down if individuals
behave selfishly; punishers thereby discipline the selfish for the good of the group.
23


cooperation in repeated interactions (Boyd and Richerson 2005, Gintis et al. 2003).
Alternatively, altruistic punishment, or strong reciprocity, is built on the principle of
reciprocity as discussed above, but is more difficult to explain in terms of self-
interest (Gintis et al. 2003).
Strong reciprocity is a form of altruism because it benefits the group at large
while the punisher suffers costs. Both ethnographic and empirical evidence show that
people punish defectors not just in repeated interactions, but also in one-shot
encounters (Boyd and Richerson 2005, Fehr and Gachter 2002). Strong reciprocators
inflict various forms of retribution, including refusal of future cooperation; physical
attack; social ostracization through rumors, gossip, or imprisonment; and denial to
resources, territory, or mates (Boyd and Richerson 2005). Results from ultimatum
games (See Section 2.4.3) and public goods experiments9 show that when costly
punishment is allowed in game theory experiments, cooperation does not decline;
and in games of anonymous pairs, cooperation actually increases when punishment is
allowed (Bowles and Gintis 2002a, 2002b, Fehr and Gachter 1998, 2002). In a
repeated public goods experiment, Fehr and Gachter (2002) found that the threat of
punishment deterred defection while punishment itself acted to reform defectors in
subsequent rounds, thereby increasing the prevalence of cooperation (Fehr and
Gachter 2002). Bowles and Gintis (2002b) argue that the motivation to punish is
stronger when the identification of the group is known (even if individuals are
anonymous), so that strong reciprocity becomes stronger when group stability is
high. Thus retributive sanctions may explain why cooperation can be evolutionarily
stable among large groups of unrelated individuals (Boyd and Richerson 2005).
Building on Gintis (2000), Gintis et al. (2003) assert that strong reciprocity
itself is an evolutionarily stable strategy and that even a small fraction of strong
9 Participants are given an endowment and may contribute any amount to a group project, earning a
percentage of the total project fund. Defecting, or keeping all the money, seems to be the most
profitable strategy, but when all members donate to the group project, joint gains are maximized.
24


reciprocators could invade a group dominated by egoists and proliferate. The authors
acknowledge the explanatory importance of inclusive fitness and reciprocal altruism
for the inception of altruistic cooperation, but assert that once structured social
interactions are established, strong reciprocity prevents defection, or the pilfering of
benefits from the cooperating majority (Boyd and Richerson 2005). Thus strong
reciprocity may further explain the evolutionary success of the human species via
large-scale and anonymous cooperation, and may be an example of gene culture
coevolution via multi-level selection.
Retribution in altruistic punishment is costly to the punisher but beneficial to
the group. Over time, punishing the defectors and reforming them towards
cooperation not only deters the defecting individual and protects the group as a
whole from that defector, it also threatens other potential defectors and deters them
from bad behavior (again, threat deters defection, and punishment enhances
cooperation). So, long-term benefits outweigh the short-term costs of punishment.
When punishers are common, defectors are selected against because they are
punished. Selection thus favors punishment, though the resulting cooperation may
not equalize the costs to the individual punisher.
Altruistic punishment is interesting because 1) cooperation may be possible
in much larger groups than with mere non-policed reciprocity; and 2) punishers pay
the cost to essentially provide a public service for the good of the group. Because of
this cost however, selection should favor cooperating non-punishers\ individuals
who cooperate in exchanges, but who do not incur costs for policing the behavior of
non-cooperators. These second order free-riders may exploit punishers if the cost of
punishment is high (Henrich and Boyd 2001, Henrich et al. 2004) so that they get
higher payoffs than the punishers do. Though the existence of second order free-
riding shakes the evolutionary stability of punishment (Henrich et al. 2004),
25


...the payoff disadvantage of altruistic cooperators relative to
defectors is independent of the frequency of defectors in the
populations, whereas the cost disadvantage for those engaged in
altruistic punishment declines as defectors become rare because acts
of punishment become very infrequent. Thus when altruistic punishers
are common, individual level selection operating against them is
weak. [Boyd et al. 2003: 3531]
In other words, altruistic punishment becomes stable when punishers are common,
thereby effectively limiting the number of defectors and thus the need to punish. This
in turn limits the impact of second-order free-riding. Cooperators have a higher
fitness than defectors if punishers are common enough so that the cost of being
punished is greater than the cost of cooperating. Thus, everyone is better off when
punishment exists, yet there are no incentives for any one individual to punish (Fehr
and Gachter 2002). Punishment of free-riders then becomes a second-order public
good, benefiting group members in the future if they amend their behavior.
Proponents of altruistic punishment (Boyd et al. 2003, Henrich and Boyd
2001, Henrich et al. 2004) assert that a cultural evolutionary model based on group
selection helps explain the evolution of punishment among humans (Boyd and
Richerson 2005, Gintis and Bowles 2003, Sigmund et al. 2002). At the core of the
model is the assertion that humans possess a unique form of social learning, namely
the copy-cat disposition that allows learning primarily through two innovative
mechanisms: 1) conformist transmission, or the copying of frequently observed
behavior acted out by the majority; and 2) payoff biased transmission'0, or the
copying of particularly successful behavior. Supported by empirical evidence from
psychology, these types of social learning allow humans to essentially leapfrog other
animalsthat must learn behavior through parental observation and experiential
trial-and-errorinto more efficient and adaptive behaviors. 3) The third mechanism,
punishment, acts as a behavior regulator to discipline non-conformers (Henrich et al. 10
10 Also called prestige-biased transmission (Henrich et al. 2004).
26


2004), while the fourth and final mechanism, 4) normative conformitythe drive to
match behavior to that of peersthen arises so that despite personal intention or
beliefs about the worth of the behavior (and therefore contrary to conformist
transmission) the behavioral majority rules. Altogether or individually, the four
mechanismsbimodal cultural transmission, punishment, and normative
conformityhelp create and maintain behavioral equilibria (like cooperation) not
allowed in strictly genetic evolutionary processes. These behavioral equilibria also
allow group selection to attain greater importance11,12. Once common, cultural group
selection may facilitate the spread of cooperative behavior to non-cooperative groups
as members of these groups 1) observe the higher productivity of cooperative groups
in either resource allocation or in militaristic domination; and as they 2) imitate, in
increasing frequency, individuals with higher payoffs that belong to cooperative
groups.
Once cooperation and punishment are established through this cultural
evolutionary process, prosocial genes can then proliferate because they will increase
fitness by making individuals less likely to befall punishment. In this genetic
cultural co-evolutionary, multi-level process, rapid cultural changes within-groups
can drive group equilibria, where they may remain stable until between-group
selection favors an alternate strategy. Henrich (2004) argues that between-group
selection will favor prosocial groups because they out-compete groups predominated 11 12
11 The authors (Henrich and Boyd 2001, Henrich et al. 2004) argue that group selection, the whipping
boy of many anti-sociobiology anthropologists and geneticists, is not an entirely separate process from
individual or natural selection and is a useful concept in order to emphasize the interaction of genes
and environment, especially when addressing cultural evolution. Per their definition, genetic group
selection comes about when natural selection acts on differences in gene frequencies between groups
and overtakes within-group forces, to select for a different equilibrium than that selected by within-
group forces acting on individuals alone. Cultural group selection works in the same way on learned
behavior.
12 An alternate view is that the concept of group selection unites the previously disparate theories of
inclusive fitness, reciprocity, and game theory (Wilson 1983) by helping explain (anomalous)
behavior that limits each theory alone.
27


by non-cooperators. When this occurs and cultural selection favors prosocial
phenotypes, selection of the prosocial genotype becomes possible.
The unique ability to internalize culture described above permits humans to
create and follow social norms. When norms are subverted or ignored, negative
emotionsresentment, angerdrive those who feel wronged to take action, often
sanctioning the norm-subverter (Bowles and Gintis 2002a, 2002b, Boyd et al. 2003,
Fehr and Fischbacher 2002a, 2002b, 2004a, 2004b, Fehr et al. 2002, Fehr and
Gachter 2002, Gintis 2000a, Henrich and Boyd 2001). These negative emotions
allegedly drive punishment in both repeated and one-shot interactions, as strong
reciprocity acts as a powerful norm enforcer. Moreover, punishment ostensibly
produces shame in the defector because the punished tend to cooperate after
punishment13. Fehr and Gachter (2002) emphasize the emotionality of this process,
where punishment is a vengeful end, not a rational means toward reform on behalf of
public well-being.
In sum, altruistic punishment, or the costly punishment of defectors without
any material gain, is driven by the proximate mechanism of negative emotion toward
defectors (Fehr and Gachter 2002). Both laboratory-controlled experiments and
ethnographic data show that players punish defectors not just in repeated
interactions, but also in one-shot encounters (Boyd and Richerson 2005, Fehr and
Gachter 1998, 2002, Fehr et al. 2002, Henrich 2000, Henrich et al. 2001, Henrich et
al. 2004, Tracer 2003, 2004). Including punishment in the explanation of large-scale
and one-shot cooperation helps to fill the gaps in the theories of reciprocal altruism,
inclusive fitness, and indirect reciprocity (Bowles and Gintis 2002a).
Though altruistic punishment does well to explain one-shot anonymous
behavior in games, is it a plausible theory for reality? Though modem military, legal,
13 I contend however, that prior defectors may not be reformed toward cooperation because of shame,
but because the costs of continuing to defect are too great when defection will probably be punished.
Thus future cooperation after punishment may again be explained in terms of self-interest.
28


and political situations may be isolated, one-shot events (Holt and Roth 1997), how
often did our ancestors interact under utterly anonymous and non-repeated
conditions? Can results from sterile laboratory experiments be extrapolated to
realistic conditions?
With the exception of inclusive fitness, the argument could be made that the
other theories for the evolution of altruistic cooperation explicate stable conditions
once established without offering up explanatory mechanisms for the viability of
such strategies in environments dominated by egoists (Axelrod and Hamilton 1981).
It seems most likely however, that prosociality among humans is far too complex to
warrant a simple, parsimonious theoretical explanation. It is feasible that natural
selection has acted on environmentally-contingent stable strategies throughout our
evolution, including some or all of the above theories (Henrich et al. 2004). This
complicated history of altruistic evolution is compounded by the fact that altruistic
behavior is extremely difficult to assess, especially among nonhuman animals
because of lack of observational data (Brosnan and de Waal 2002, Chapais et al.
2001). Other issues include 1) the difficulty of assessing costs and benefits,
especially when the two are of different currency or relative value to different
individuals based on age, size, or rank; 2) the confounding effects of reciprocal
altruism and kin selection, that is, the difficulty in assessing genetic relatedness of
individuals; and 3) temporal issues that make it difficult to determine whether and
when returns are made.
To bolster debate about the evolution and maintenance of altruism, further
empirical testing in varied environments is needed. In recent decades, experimental
methods devised in the field of economics have been employed to test whether or not
humans are innately selfish utility maximizers, with the aim of further understanding
human altruism.
29


2.4 Behavioral Economics
2.4.1 A Description and History
According to Camerer (1999, 2003), behavioral or experimental economics is
a reunification (for they used to be more closely cooperative) of psychology and
economics in order to explain human behavior. Understanding the cognitive
processes that drive humans informs and augments the mathematical rigor of
economics. Proceeding from a standard economic approach, and based on Nashs
idea that there exist equilibria of optimal strategies, behavioral economics and game
theory transcend the limited principles of human rationality that dominate economic
theory toward alternative and more realistic theories of behavior offered by
psychology. While analytical game theory is a mathematical modeling of what
players with varied cognitive abilities will do in certain situations (the game),
behavioral game theory is concerned with what players actually do in games; it thus
involves more qualitative concerns such as actors intentions and emotions.
Experimental economic games test the predictions of game theory that individuals
are innately self-interested in a controlled environment (Romp 1997). Biologists,
philosophers, political scientists, and social scientists use game theory to test
theoretical predictions about other social preferences and tastes that influence human
behavior (Camerer 2003, Gintis 2000a, Gowdy and Seidl 2004, Holt and Roth 2004,
Kreps and Rubinstein 1997).
Shapley (1953) defines a game as a set of rules that govern the possible
actions of players: participants are presented with the games rules and then
strategize accordingly. Person A anticipates what Person B will do; Person A also
anticipates what other players will think of his own actions. In repeated games,
players may base strategies on previous interactions so that results are different than
in one-shot games. Payoffs represent utility, the economic proxy for biological
fitness, so that stability of choice depends on the relative fitness (utility) outcome of
a games strategies (Fishman 2003). Results from game experimentation produce a
30


sort of mathematical x-ray of social situations and the decisions that drive them
(Camerer 2003).
Game theory has its roots in the work of von Neumann and Morgenstem in
the 1940s (Holt and Roth 2004, Kreps and Rubinstein 1997). After World War II,
formal game theory rapidly expanded, culminating in a number of advances. The
most important of these is probably the Nash equilibrium, an extraordinary one-page
proposition published in 1950 that earned Nash the Nobel Prize some 44 years later
(Camerer 2003, Holt and Roth 2004, Kreps and Rubinstein 1997). In this paper,
submitted while still a graduate student, Nash proposed the solution to how rational
players will behave in ^-person games when players have a finite set of strategies,
including mixed strategies,14 that correspond to a payoff to each player. Nashs
theory predicts that after adjusting their strategies, players will reach a stable
equilibrium wherein no player can benefit from a change in strategy. In other words,
the theory predicts that each player will choose the best, utility-maximizing strategy
in anticipation of other players strategies (Camerer 2003). Should all players
announce their strategies simultaneously, none would want to change his choice
(Holt and Roth 2004). With the Nash equilibrium at its core, from the 1970s on game
theory gained recognition and moved from its esoteric roots to join the mainstream
language of economics (Kreps and Rubinstein 1997).
Game theory is given the credit for bringing experimental methods into the
field of economics as it lays out testable theoretical predictions about strategic
behavior (Holt and Roth 2004). Experimental economists put real people in
controlled laboratory settings, to observe their behavior when playing for real
(usually monetary) payoffs. Game theory is powerful because of its generality and
mathematical precision. Games are usually repeated to understand the proclivity
toward equilibrium. There is a growing body of literature in experimental economics
14 Mixed strategies are probability distributions over decisions; that is, a player may sometimes
choose different actions, for example a poker player who sometimes bluffs (Holt and Roth 2004).
31


both supporting and contrary to equilibrium predictions. The predictive assumptions
of the Nash equilibrium are useful both when accurate and when they do not predict
game behavior, because they highlight the existence of prosocial proclivities.
2.4.2 Homo oeconomicus
Economists define pure altruism as the act of increasing another individuals
utility at a cost to oneself (Camerer 2003). A fundamental assumption in game theory
stems from two mainstays of canonical economic theory, that players (i.e., humans)
are both 1) highly rational and 2) inherently self-interested utility maximizers. While
the latter echoes the selfishness axiom of standard evolutionary theory, the former
endows individuals with more self-awareness and individual autonomy than
evolutionary theory assumes. The rational axiom of economic theory assumes that
individuals will evaluate every strategy available to them in reality, and by extension,
in experimental games. After such evaluation, players in a game will choose the
strategy yielding the most desirableor utility maximizingstrategy according to
self-interested tastes and preferences (i.e., they will tend toward the Nash
equilibrium, or subgame perfect equilibrium that exists when there are multiple
strategies available) (Henrich 2000, Nash 1950a, Shapley 1953). However, empirical
results consistently show that actual behavior is inconsistent with these assumptions
(Henrich 2000).
Another assumption of game theory is that games are cooperative affairs
(Nash 1950a, 1950b, Shapley 1953), though in many cases (and indeed in the game
described in this thesis), they are not. In anonymous non-cooperative games, players
do not know the identities of one another and are not allowed any communication. So
players make their rational decisions without consultation with other players,
highlighting the individualistic nature of decisions (Gowdy and Seidl 2004, Roth
1997). However, Roth (1997) argues that decisions of players are also mutually
interdependent, because the welfare outcome of (at least) one individual depends on
32


the actions of another. This dependence sets the stage for strategy, and encourages
(or forces) players to consider their actions in reference to outcomes for other
players, as well as to anticipate their own outcomes at the mercy of other players.
The acknowledgment of mutual interdependence15 distinguishes game theory from
traditional economic theory.
A final assumption, again stemming from canonical economics, is that culture
is static or at least extremely slow to change (Gowdy and Seidl 2004). Economics
can therefore handily dismiss culture and assume the universal homogeneity of
human behavior in economic contexts.
This model of universal economic man, termed Homo oeconomicus,
mirrors the biological model of man as innately self-concerned with fitness and
reproductive success. Both traditional biology and economics limit individuals to
fitness- or utility-maximizing endeavors, leaving little room for cooperation or
altruism, unless as a means to gaining an advantage over some other individual
(Gowdy and Seidl 2004).
Ethnographic and empirical evidence from anthropology, biology,
psychology, and behavioral economics shows that humans are not driven solely
toward utility maximization. Despite its existence however, altruism remains outside
standard welfare economic models. These varied but incomplete models fail to
acknowledge the complicated nature of social interactions that involve competition
and cooperation. In fact, game theory experiments often miss this as well, as
laboratory experiments usually consist of isolated dyads acting in simulated
conditions, with few available options and little or no communication between
15 Mutual interdependence is ignored in neoclassical economics because of assumptions about markets
and market failure. Mutual interdependence also permits the condition of Pareto inefficiency, or the
condition where no actors welfare can be ameliorated without damaging the welfare of another actor
(Debreu and Scarf 1963), a taboo condition in neoclassical welfare models which emphasize
individualism and rational choice in producing ideal outcomes.
33


actors. Nonetheless, game theory provides an important starting point from which to
evaluate results from behavioral economic experiments.
Debate about the applicability of predictions and generalizations of economic
and game theory echoes the formalist-substantivist debate in economic anthropology
(Burling 1962, Cook 1966, Dalton 1969, LeClair 1962, Polanyi et al. 1957). In 1957,
Polanyi et al. described the dichotomous meanings of economic. The substantive
definition of economic refers to the fact that individuals depend on both nature and
peers for survival. That is, continuous interactions with his social and physical
environment supply an individual with the means to satisfy his material wants and
needs. The formal (and conventional) definition of economic refers to the logical and
rational character of the means-end relationship described above, in which
individuals choose between alternate means, some of which are better than others in
the satisfaction of material wants. Scarcity is implicit in the formalist view. There is
a finite supply of available resources for the satisfaction of wants; so while scarcity
produces deficits in alternate means, and forces the making of rational choices in the
formalistic view, scarcity need not be a factor in the substantive view.
The formalist-substantive debate is an argument of whether the conventional,
formalist view is helpful in the social sciencesspecifically in the examination of
non-Westem, primitive culturesas it is substantiated on the elemental
assumptions of neoclassical economics (e.g., rational action, market behavior, and
pricing). Substantivists (Dalton 1969, Polanyi et al. 1957) tend to emphasize
induction and cultural relativism, and warn against asking questions based on our
own (Western) economy. Furthermore, substantivists generally believe that the
difference(s) in Western and non-Westem economies is one of kind while formalists
believe the difference is one of degree. Formalists (Burling 1962, Cook 1966,
LeClair 1962) emphasize deduction and the universality of economies, insisting that
our own economy is a good reference point from which to compare other types of
economies.
34


Likewise, it could be argued that game theory, based on conventional
economic theory, is not only unrealistic in industrialized nations (because it
presumes that controlled, anonymous, laboratory interactions approximate reality),
but it is especially inapplicable to studies in small-scale societies (because of the
latter, in addition to presumptions about the universality of market economies). Like
formalists however, proponents of game theory argue that theoretical groundings are
a good starting point and greatly simplify analysis. The heuristic value of theory is
not only made manifest when theory accurately predicts behavior, but when it
illuminates anomalous behavior deserving of further study.
2.4.3 Games: The Prisoners Dilemma, Dictator Game,
and Ultimatum Game
The prisoners dilemma (PD), a model for political, business, and biological
interactions (Bartholdi et al. 1986), is a static gameone in which all players act
without information about other player actionbased on the scenario16 of two
suspects being arrested for the same crime without sufficient evidence to firmly
convict either unless at least one confesses (Holt and Roth 2004, Romp 1997). If
neither confesses, both will be convicted of a minor offense and sentenced to one
month; if both confess, they will each be sentenced to six months; if only one
confesses, the confessor will be released while the other will be sentenced to nine
months, that is, six months for the crime and three months for obstruction of justice.
In the experimental game version of the prisoners dilemma, the two players
are given the option to cooperate or defect with the monetary payoffs for each action
loosely based on the punishmentreward scenario described above (Bartholdi et al.
1986, Holt and Roth 2004). Table 1 summarizes an example of such pay-offs.
16 The narrative of the Prisoners Dilemma was devised by Albert Tucker, Nashs thesis advisor (Holt
and Roth 2004).
35


Table 1 Example of a Prisoners Dilemma Payoff Scheme
Player A
Cooperate Defect
Cooperate (80,80) (0,100)
Defect (100, 0) (35,35)
There are four possible outcomes: (cooperate, cooperate), (cooperate, defect), (defect, cooperate), or
(defect, defect). Payoffs are in the format (Player A, Player B).
The pure Nash equilibrium strategy is mutual defection; however, if both players
cooperate, they are both better off than if they both defect (Romp 1997). If players
are indeed selfish egoists, the preferred or most profitable outcome for each player is
defection while the other player cooperates.
As mentioned above in the section on reciprocal altruism, through several
rounds of computer simulation, Axelrod and Hamilton (1981) showed that not
always defect (ALLD), but tit-for-tat (TFT) is the prevailing and most successful
strategy in iterative rounds of the prisoners dilemma. Thus the computer uses a
strategy derived from past experience of opponent cooperation or defection
(Bartholdi et al. 1986) instead of selfishly always defecting. Bartholdi et al. (1986)
extend the findings of Axelrod and Hamilton (1981) to argue that not just
individuals, but corporations and even nations conduct behavior using a TFT
strategy. Once established, TFT is extremely robust and resistant to intrusion by
mutant strategies. Leaving computer simulations behind, the following examples
come from economic games performed among actual individuals.
Empirical results from iterative rounds of prisoners dilemma games
consistently demonstrate that with time, individuals move from a high level of
36


cooperation toward more frequent defection. For example, players usually cooperate
40-50% of the time in initial rounds, but cooperation drops to around 20% with
repeated play (Romp 1997). Romp (1997) argues that this demonstrates increased
game experience influences participants to choose cooperation less and less, thus
tending towards the Nash equilibrium (ALLD). As Axelrod argues (1984), a more
accurate statement might be that players adopt a TFT reciprocal strategy. They
choose cooperation in the first round and then base subsequent decisions on
experience. However, one-fifth of the participants still chose the cooperative strategy
after several rounds, so it seems that the Nash prediction only receives limited
support (Romp 1997), as does the assumption that individuals are rational (or will
play rationally) and that they know their opponents are rational as well. Potential
explanations for such altruistic behavior assert that players get additional utility from
cooperation in one of three forms (Andreoni and Miller 1993). In (1) pure altruism
the player is concerned not only for himself, but for the welfare of the other players,
so that additional utility is received due to the greater gain for the other player;
because of (2) duty the player cooperates from moral obligation, so that additional
utility is received from cooperation; in (3) reciprocal altruism the individual receives
extra utility when both players cooperate; this form is sometimes termed warm
glow, as mutual cooperation is supposed to convey pleasure for both players17. In
other words, payoffs do not just consist of the money doled out by the experimenter,
but include intangibles. This rationalization suggests that seemingly irrational
behavior may in fact be rational and utility-maximizing, when utility is considered in
non-monetary payoffs.
17 Though previously discounted, this idea may receive additional support in the near future from
neurological studies of the basis of altruistic punishment that assert that not strategy, but good
feelings drive the punishment of norm-violators; in other word, altruistic behavior, including
cooperation and the punishment of defectors is psychologically rewarding, and may thus be
evolutionarily-based (Fehr and Rockenbach 2004, Quervain et al. 2004).
37


Another simple and frequently performed game is the dictator game (DG). In
the DG, Player A, the dictator, is given a sum of money that he may divide between
himself and Player B. Player B is inert and merely receives whatever, if any, payoff
Player A gives him18. In non-anonymous games, the average amount shared is 50%;
when player identity is kept anonymous, the amount shared drops to 36%, a
percentage much higher than the expected offer of 0% should players act according
to selfish motives. Even more astounding are results from the ultimatum game, a
variation of the DG that adds the element of punishment by Player B.
The ultimatum game (UG) consists of an exchange between two players, with
a potential gain for both if the players can agree on how to divide up the sum
(Camerer 2003, Fehr and Gachter 1998, Gowdy and Seidl 2004, Henrich 2000,
Nowak et al. 2000). Player A, the ultimatum-giver, proposes a take-it-or-leave-it
offer of how to divide up the endowment with Player B. If Player B agrees, Player B
receives the proposed amount while Player A takes the remainder; if Player B does
not accept the proposal, neither player enjoys any gain19 20. There is no negotiation.
Typically, like the PD and DG, payoffs are given in real money and the players are
anonymous to everyone except the experimenter Game theory (via economic
theory) predicts that players (or more generally humans, H. oeconomicus) are selfish
and rational so that in the one-shot ultimatum game, Player A will offer as little as
possible and Player B will accept any non-zero offer (Camerer 2003, Henrich 2000,
18 Because only one player has the opportunity to make a decision and act, some argue that the DG is
not really a game.
19 For example, Player A is given $10, and must offer some amount, x, to Player B. If Player B agrees,
he accepts x, while Player A keeps $ 10-x. If Player B rejects the offer, neither player receives any
money.
20 Some experimental economists (Hoffman et al. 1994) arrange the game so that even the
experimenter is anonymous, to control for experimenter-influence of player behavior. Hoffman et al.
(1994) highlight the potential for costly signaling by players that may promote altruistic behavior (i.e.,
the player wants to appear generous to the watchful experimenter, and thus proposes a higher offer,
despite his true sentiment).
38


Gintis et al. 2003, Gowdy and Seidl 2004, Nowak et al. 2000, Sigmund et al. 2002)
because something is better than nothing. However, Tracer (2003) predicts that
Player B will not accept anything less than a 5050 split because an unequal split
might produce an imbalance in relative fitness, under the simplifying assumption that
each unit of payoff is convertible to a unit of fitness.
UG experimentation with university students in numerous culturally and
geographically diverse developed countries seems to disprove the above economic
theoretical assertion that players are inherently selfish. The modal Player A offer is
consistently 50%, and the mean offer between 40% or 50%; offers of less than 20%
are refused about half of the time regardless of sex, age, and degree of anonymity
(Camerer 2003, Roth et al. 1991, Henrich 2000, Gowdy and Seidl 2004, Nowak et al.
2000, Roth et al. 1991, Sigmund et al. 2002). Arguably, the same results occur when
the stakes are higher21, that is, in low-income developed countries, where the stake is
equivalent to several weeks work of wages (Cameron 1995, Fehr and Gachter 1998,
Fehr et al. 2002, Gowdy and Seidl 2004, Henrich 2000). It seems, then, that players
are motivated by some sense of fairness independent of their own payoff; or that they
are willing to (altruistically) decline offers of free money in order to punish other
players for their unfairness. Fehr and Gachter (1998) assert then, that not H.
oeconomicus, but Homo reciprocans, is the more accurate taxonomic designation of
players (Sigmund et al. 2002). Rejections of non-fair splits, or non-5050 offers,
may be an example of strong reciprocity (altruistic punishment), where Player B
reciprocates the unfair behavior doled out by Player A even though retribution comes
at a cost to himself. The important question becomes the origin of this principle of
fairness. As mentioned above, Tracer (2003) argues that fairness (equal split) is
21 This does seem to work with relatively higher stakes in under-developed countries. However, Telser
(1995) shows that when the stakes are very high, rejection is very low or non-existent even if offers
are as low as one percent. For example, if the original endowment is $1,000 and the offer is $100,
Player B is very unlikely to reject. Thus, following the law of demand, the split approaches extreme
inequality with increasing endowments.
39


congruent with the other-regarding, relative fitness-maximizing predictions of
evolutionary theory.
Recent behavioral economic experiments by eleven anthropologists in Africa,
the Amazon, Papua New Guinea, Mongolia, and Indonesia demonstrate that in these
small-scale societies (foraging, horticulturalist, nomadic herding, and small-scale
agricultural societies), results are very different from those in developed countries
where games were typically played with university students (Bolton et al. 1998). In
an anonymous UG played among the Machiguenga of the Amazon, Player A offered
very little (mean offer: 26%; mode offer: 15%) while Player B consistently accepted
almost every offer, even those less than 20% (Henrich 2000). These offers and
rejection rates are much lower than those found among university students.
Conversely, among the Au of Papua New Guinea, Player A sometimes offered more
than 50%, some citing fear of village turmoil should any (too-low) offer be refused
(Henrich et al. 2001, Tracer 2003, 2004). In the above examples, the experimenters
concluded that culture was the primary factor determining behavior, including
expectations of proper offers and beliefs about fairness. In the Papua New Guinea
case, over-sharing was explained by cultural rules about generosity, exchange, and
reciprocity, such as those that prescribe that hunted game cannot be consumed by the
hunter and must be shared. However, responders in Papua New Guinea rejected
hyper-fair offers as well, allegedly due to cultural rules about competitive gift-giving.
These rules dictate that taking a large sum incurs even greater future obligation.
Indeed, the Au tended to reject both excessively generous and low offers equally.
Henrich (2000) asserts that the Machiguenga felt neither an obligation to divide the
endowment equally, an expectation to receive an equal share, nor an innate desire to
punish an unequal division. In other words, they felt the modal offer of 15% was
fair. In sum, results from these experiments in small-scale societies show that 1) the
canonical economic model of self-interest is not supported; 2) individual-level
demographic and economic variables do not explain behavior within or between
40


groups; 3) there is great behavioral variability cross-culturally, but behavior is
explained in part by degree of market integration22 (hence the difference in results
between university students in developed nations and in participants from small-scale
societies); 4) local economic patterns are generally congruent with degree of
cooperation (i.e., economies of scale in production) and/or punishment. It seems
then, that different cultures produce different standards of fairness, often consistent
with degree of anonymous market exchange (Camerer 2003, Henrich et al. 2001,
Tracer 2004). Also, with an increase in market integration and Westernization,
cultures seem to have sharing norms that produce more equal splits. In other words,
in cultures with the most market integration, offers seem to be the least selfish
(Camerer 2003). But is it that market experience creates regulating norms about
fairness and equal division, or that the proclivity toward even distribution produces a
favorable market environment?
Offers and rejections in cross-cultural ultimatum games may be said, then, to
be a sort of language portraying cultural nuances. For both developed and non-
developed countries, UG results suggest that many players play by cultural rules
according to what is locally fair (Clark and Sefton 2001, Eckel and Grossman 1998,
Schroeder et al. 2003). These may include reaction to unfairness 1) in outcome
(Bolton and Ockenfels 2000, Fehr and Schmidt 1999), 2) in intent (Rabin 1993), or
3) in both outcome and intent (Chamess and Rabin 2002, Falk and Fischbacher 1999,
Falk et al. 2003). The first is based on the assumption of a utility function, and refers
to equitable outcomesin game theory, payoffsas based on some measure of how
close an actual outcome or offer approximates a fair reference pay-off. Theories of
fairness in outcome may help explain the one-shot prisoners dilemma, public goods,
gift-exchange, and ultimatum game rejections (Diekmann 2004). The second and
highly rational theory assumes that stakes matter, and refers to an evaluative
22 Market integration refers to an index including existence of a national language, existence of a labor
market and fanning for cash (Camerer 2003).
41


response in kind for intentional22 behavior; that is, eye for an eye, whether the eye
is benevolent (fair) or vengeful (unfair). This theory may again explain ultimatum
game rejections. The third argues for the importance of fairness in both intent and
outcome. Still others (Chamess and Rabin 2002) claim that punishment may be
explained by competitiveness of individuals who prefer to have their payoff be as
high as possible relative to other payoffs. This theory sounds strikingly similar to that
of other-regarding relative utility maximization (Tracer 2003).
It has been argued that humans universally value the emotional and moralistic
idea of fairness (Sigmund et al. 2002). Sigmund et al. (2002) propose that humans
evolved the emotions at the heart of this situation from life in small groups, where
behavior that not only benefited individuals but that benefited the group in the long-
run was favorable. The authors suggest that fairness is a sort of self-conscious
behavior, because others (in our group) are watching and remembering what we do,
and thus may predict what we are likely to do in the future. In other words, we all
have and must maintain a reputation. One-shot encounters were infrequent
throughout human evolution, so that individuals tend to respond angrily to defection
in an effort to nurture both self-esteem and reputation, as an individual that will not
tolerate being cheated. While it could be argued that humans indeed have something
called good character, and that they truly enjoy helping and sharing with others,
surely there must be a biological component to fair behavior (Sigmund et al. 2002).
Sigmund et al. (2002) argue that altruism, and other social emotions like friendship,
guilt, resentment, and shame, help us negotiate (ultimately biological) success in
complex social networks.
Results from dictator and ultimatum games as well as more complicated
game experiments (public goods games, gift-exchange) consistently show that 23
23 Trivers (1971) also pointed out the problem of intent when defining altruism, that is, whether to
identity altruism by motive or by behavior regardless of motive.
42


humans are not primarily selfish. Theoretical economic and evolutionary views of
behavior, then, are poor predictors in lived realities where cultures and social norms
play a much bigger role than previously allowed in the traditional view of H.
oeconomicus, the rational and self-interested actor (Fehr and Gachter 2002, Gowdy
and Seidl 2004, Henrich et al. 2001). This suggests there is a need for a new
behavioral economics (Gowdy and Seidl 2004).
In order to further the debate about the roles of altruism, fairness, and justice
in game behavior, this thesis reports the results of a modified dictator game. As in the
dictator game, a proposer decides how he and an anonymous second party will divide
an endowment. Then an anonymous third-party is given an endowment equal to half
of that divided between the other two players; he is also give the opportunity to usurp
the power of the dictator and change the proposal. He may sanction the proposer,
compensate the recipient, or both punish and compensate, but all at a cost of a
percentage of his endowment; or, he may take his full endowment and leave the
payoffs as they stand. From a rational and selfish perspective, third-party punishment
or compensation is altruistic and unjustified: it reduces the utility of the third-party.
2.4.4 Weaknesses of Game Theory
Ideally, game theory has the potential to provide a holistic and realistic
window into economic behavior with the potential to elucidate understanding of the
evolution of that behavior (Gowdy and Seidl 2004). However, game theory has been
criticized for its simplicity, abstraction, lack of real-life haggling and bargaining,
lack of face-to-face contact, lack of a feed-back loop, rigid structure that disallows
mutation in the form of player ideas, limited time for making evaluative decisions,
lack of realistic context, and typical limit to one-shot non-repeated encounters
(Camerer 2003, Gowdy and Seidl 2004, Romp 1997). For example, the prisoners
dilemma has been criticized for predisposing players toward selfish action because
players may not communicate, cheating is rewarded, and the punishment of free
43


riders is not possible. Such lack of autonomy and communication denies players of a
primary and perhaps unique human attribute (de Waal 1996). Nonetheless, game
theory remains attractive in the social sciences for the potential of quantification of
the intractable subject of human behavior (Sigmund et al. 2002).
In summary, simple experimental economic games like the ultimatum game
test game-theoretical principles, and are useful to understand what people think about
the allocation of resources to both themselves and their peers (Camerer 2003, Fehr
and Gachter 1998, Sigmund et al. 2002). Game theoretical experiments consistently
show that when given the opportunity, players will respond in kind (Rabin 1993),
rewarding those who are generous and punishing those who are not (Fehr and
Gachter 1998). However, they also tend to play games according to socially and
culturally informed rules of fairness. In addition, since most experiments are
anonymous, reciprocity and punishment seem to apply even when people do not
know with whom they are dealing.
2.5 Conclusion
Since the publishing of Adam Smiths24 Wealth of Nations, standard
economic theory has assumed that all humans are in the business of maximizing their
self interests (Balasko 1988). In the jargon of economics, when consumers are faced
with the choice between two commodities, they are sentient and rational enough to
compare the two and choose the preferred commodity based on preference (Balasko
24 .. .but man has almost constant occasion for the help of this brethren, and it is in vain for him to
expect it from their benevolence only. He will be more likely to prevail, if he can interest their self-
love in his favour, and show them that it is for their own advantage to do for him what he requires of
them....Whoever offers to another a bargain of any kind proposes to do this. Give me that which I
want, and you shall have this which you want, is the meaning of every such offer; and it is in this
manner that we obtain from one another the far greater part of those good offices which we stand in
need of. It is not from the benevolence of the butcher, the brewer, or the baker, that we expect our
dinner, but from their regard to their own interest. We address ourselves not their humanity but to
their self-love, and never talk to them of our own necessities but of their advantages (Smith 1868:6-7)
(emphasis added).
44


1988). And we might infer that this preferential choice should be based on the drive
to maximize self-interest. This idea mirrors the assumption of canonical evolutionary
theory that all organisms are self-regarding, with the aim of maximizing individual
fitness and subsequently reproductive success.
According to the evolutionary theories of kin selection and reciprocity,
cooperative behavior is favored by selection only when cooperators are more likely
to interact with other cooperators; kin relationships seem to provide the most likely
source of this (non-random) interaction (Boyd and Richerson 2005). Human
eusociality may thus be explained by the combination of this reciprocity and the
(perhaps unique) human ability to use our big brains to recognize a large number of
individuals and to remember a large number of (historical) social interactions. These,
coupled with punishment as a disincentive, may be the keys to understanding human
cooperation.
Hamiltons idea of kin selection is the most convincing model for the
inception of altruism. Like primate groups, the first groups of humans were primarily
made up of kin, with some mixing of lines to avoid inbreeding. Cooperation first
arose, as seen among non-human primates, as a reproductive strategy for group-
living, to avoid predators and to permit the best utilization of resources. As
intelligence increased and social organization grew more complicated, cooperation
too grew increasingly more complicated to ensure social bonds. Cooperative ties and
reciprocal acts began to extend to non-kin, as encouraged by previous positive
experiences with relatives. As interactions increasingly took place between distantly-
and non-related individuals, natural selection favored the ability to both remember a
large number of past interactions and to discern honesty. Negative reciprocity, also
termed altruistic punishment, naturally flowed from positive reciprocity, as a
deterrent for defectors. The theories of both direct and indirect reciprocity and of
altruistic punishment, explicate what happens once cooperation comes into vogue,
better than they explain how cooperation could arise and evolve. From Hamiltons
45


theory, the extrapolation of the benefits of altruistic cooperation can be
synergistically extended. Reciprocity and punishment are sustaining forces after
cooperation is already in place, especially in large groups with highly specialized
division of labor. Thus kin relations first produced cooperation; punishment sustains
it among non-related individuals.
Though primatologists, biologists, and anthropologists have long dismissed it,
explorations of altruism often return to the controversial ideas of group and multi-
level selection because kin selection and social reciprocity fail to fully explain
altruistic behavior (Bradley 1999, Darlington 1972, 1978, McAndrew 2002, Smith
1976, Strier 2000). There is evidence for group selection among virus strains and
foraging ant queens (Bradley 1999). By nature, altruism occurs in groups, even if
only groups of two (Darlington 1978). If selection occurs at the level of the group,
then groups with altruists should have higher fitness than those without them (Strier
2000). But in nonhuman groups, it seems that the animals who warn others of
predators increase their chances of detection and thus decrease their fitness; in
addition, selfish individuals would pass on more genes by free-riding on such
warnings. However, if a distinction is made between individual selection, which acts
on differential fitness of individuals within groups, and group selection, which is the
same principle acting on differential fitness of individuals between groups, one
altruist surrounded by a horde of selfish individuals will have lower individual
fitness; but a group of cooperative and altruistic individuals may have higher fitness
than individuals in other groups composed of mostly selfish members. If a multi-
level hierarchy of selection pressures acts on individuals, on populations, and on
species, et cetera, and differences are heritable at each unit of selection, natural
selection could conceivably operate at multiple levels in a sort of co-evolution
(Darlington 1978, Strier 2000). Strier (2000) cites the example of a selfish group that
overexploits its niche, thereby reducing the average fitness of all group members. In
juxtaposition, cooperative members of a largely altruistic group may have lower
46


individual fitness than selfish members of other groups, but over time will persist
longer than the selfish group. Though selfish members could infiltrate an altruistic
group, if punishment becomes the norm, their selfishness can be stifled. Darlington
(1978) agrees to a degree, but argues that though altruism is a group issue, it evolves
by individual selection, is opposed by competition, costs, and inefficiency, but is
supplemented by group selection in a multi-level process. Though these issues are
very difficult to study among nonhuman animals, game theory provides an important
foundation from which to build explorations of altruistic cooperation.
Is it an evolutionary accident or even maladaptation (Fehr and Henrich 2003)
that humans cooperate, especially with strangers with whom they meet only once
(Gintis and Bowles 2003)? Is this ability the result of the uniquely human qualities of
language or cognitive ability, or the cultural tendency to have prosocial norms and
institutions that guide our social conduct toward fairness, including the tendency to
punish violators? Humans often live and work in groups of non-kin. Is cooperation a
necessary pre-condition, or by-product of this arrangement? A substantial body of
evidence supports the idea of kin selection among cooperating animals; a less robust
body of evidence supports reciprocity as an explanation of animal, specifically
primate, interactions. These theories do well to explain human interaction with kin
and individuals with whom humans repeatedly interact, but costly signaling and
strong reciprocity (altruistic punishment) carry evolutionary theories further, yet not
far enough to explain one-shot altruism. The interaction of both genes and culture are
necessary to understand altruistic cooperation in an evolutionary framework. Perhaps
the idea of multi-level selection on variability in both genes and culture provides the
key to understanding altruism (Boyd and Richerson 2005). Though existing theories
do an adequate job of explaining cooperation among kin and for non-kin who interact
repeatedly and frequently, explanations of one-shot altruism remain unclear so that
further empirical testing with the aim of refining theory is required.
47


3. Study Population
3.1 The Setting
3.1.1 Papua New Guinea
Papua New Guinea (PNG) is located just north of Australia along the Ring
of Fire and shares the second largest island in the world with the Indonesian region
of West Papua, formerly called Irian Jaya (See Map A.l). The 820 km land boundary
with Indonesia not only divides the island of New Guinea in half vertically, but
separates the two nations culturally. PNG is most often grouped into the South
Pacific islands known as Melanesia, along with its nearest neighbor to the east, the
Solomon Islands, and Fiji, New Caledonia, Vanuatu, Maluku, and the Torres Strait
Islands.
West Papua became a part of Indonesia in 1962 after the Dutch lost
administrative control (Tracer 1991). From 1885 until 1902, control of the eastern
half of New Guinea was maintained by Germany in the northern Territory of New
Guinea and the United Kingdom in the southern Territory of Papua. In 1902,
Australia took control of the southern territory on behalf of the United Kingdom
(Banks 1998); subsequently, during World War I, Australia gained administrative
control of the northern territory from Germany as well (Tracer 1991). Australia
maintained control until 1975 when the nation state of Papua New Guinea achieved
independence. PNGs geo-political situation changed once again in 1997 with the
secession of Bougainville, the largest of the Solomon Islands, following a bloody
nine-year revolt (Banks 1998).
48


Geographically, PNG boasts 5,152 km of coastline and a total area of 462,840
sq km an area slightly larger than the state of California (CIA 2005). The rugged,
tropical terrain is mostly mountainous with coastal lowlands and rolling foothills,
rendering only 0.46% of land arable for agriculture. The highest point is Mt.
Wilhelm at 4,509m. Though PNG possesses considerable natural resources including
gold, copper, silver, natural gas, timber, oil, and fisheries, the rugged terrain and lack
of infrastructurethe total length of roads in PNG is 19,600 km, with only 686 km
pavedprevent major exploitation of these resources (to the delight of travelers and
anthropologists alike). Still, 72% of export earnings come from mineral deposits
including oil, copper, and gold. As a result, pollution from mining and deforestation
are the primary environmental threats. Finally, 20% of the national budget is derived
from the $240 million in aid received annually from Australia.
In 2000, the PNG census reported a population of 5,190,786 people (PNG
NSO 2005), with current estimates (July 2005) of 5,545,268 citizens. The population
has a median age of 21 years with an average life expectancy of 65 years25 26. The
current total fertility rate is 3.96, but this figure varies greatly by region. Eighty-five
percent of the population survives on a subsistence lifestyle of gardening, animal
husbandry, and limited cash cropping (Banks 1998) including coffee, cocoa, and
increasingly vanilla. The social organization is what Henrich (2000) has called a
family-based society, meaning that extended families, or clans, predominantly
produce for themselves and do not rely on institutions for their welfare. In such
societies, anonymous transactions are almost unknown, and a high degree of
reciprocity pervades daily life.
Though English is the national language of PNG, only 1 -2% of the population
is fluent. The lingua franca of PNG is Melanesian pidgin (Tok pisin), but an
25 Including 452,860 sq km land area and 9,980 sq km water area.
26 Male and female life expectancies are 62.76 and 67.21 years, respectively (PNG NSO 2005).
49


estimated 700-800 indigenous languages are spoken in PNG, many of them unrelated
(Banks 1998, Tracer 1991). The lack of infrastructure and relative isolation
mentioned above not only currently restricts development, but helped create and
continues to maintain the cultural differences and linguistic diversity found among
groups even when the absolute distance between them is minimal.
Papua New Guinea is a constitutional monarchy with parliamentary
democracy (Banks 1998). The legal system is based on English common law. This
unitary system, a vestige of Papua New Guineas colonial past, provides that while
PNGs 19 provinces and national capital district have local legislative power,
provincial legislation may be vetoed by the creation of National Acts (Banks 1998).
The official criminal justice system of PNG is also a remnant of colonization.
Western assumptions about guilt, innocence, retribution, and individual
responsibility pervade the imposed criminal justice system, making it often at odds
with traditional ideas about principles of justice and the methods used to keep it. The
disconnect in ideas about justice, the language used to describe it, and the protracted
action of the courts in trying cases leave the judicial system largely incomprehensible
to many indigenous peoples who generally believe in swift reprisal to address violent
crime, and largely emphasize victim compensation over other punitive measures
(Banks 1998).
3.1.2 Anguganak, Sandaun Province
This study was conducted among the Au of Anguganak, Sandaun Province,
in northwest Papua New Guinea. Anguganak comprises a series of hamlets about 50
kilometers inland from the northern coast and 95 kilometers east of the Irian Jayan
border (Tracer 1991) (Map A.l). The Au occupy the southern foothills of the
Torricelli Mountain range, living at altitudes of 150 to 850m. This mountain range 27
27 Formerly West Sepik Province.
50


isolates the Au from the Pacific Ocean. The Au populate lowland tropical rain forest,
typical of an equatorial climate where temperatures vary little annually. Annual
rainfall is very high, exceeding 2.5m. Though the periods from October to March and
from April to September are classified as the wet and dry season respectively, the dry
season is not typified by drought but a reduction in overall rainfall with accompanied
drops in river and spring flow. Malaria is endemic to the area as the climate
encourages the proliferation of anopheline mosquitoes; indeed, malaria is the
preeminent cause of both child and adult mortality. Other health concerns include
dengue fever, tuberculosis, tapeworm, tropical ulcers, and scabies.
3.2 The People and Culture
3.2.1 The Au
Au not only refers to the approximately 10,000 individuals who occupy
about 50 villages in the East Au and West Au census divisions, but is also the name
of the predominant28 language spoken in the area (Tracer 1991). Villages, generally
constructed atop ridgelines, range from less than 100 to more than 500 people. The
spatial layout of the villages, constricted by the width of the ridgeline, usually
consists of several hamlets strung together by mudstone paths. Most likely the result
of missionary influence, the current housing layout and therefore sleeping
arrangement is very different from that described by Lewis (1980) and Tracer (1991).
Mens houses that used to provide sleeping quarters for all men and most boys over
the age of 10 have disappeared. Nuclear families now reside together, some in
ground-level, windowless, thatched-roof houses; but many live in more modem
28 Other languages are spoken within the East and West Au census boundaries, including Gnau (Lewis
1980), Elkei, Ghal, and Yil, which together make up an ancient and unique phylum of languages
ostensibly distinct from other languages outside of the Sepik region, and arguably may provide a link
to the original languages brought to PNG by incipient immigrants arriving from the Malay area
(Tracer 1991).
51


houses built atop stilts, some with screen-covered windows, multiple rooms, and
increasingly, corrugated iron roofs (Photographs A.l, A.2, and A.3).
Tracer (1991,2003) characterizes the Au as forager-horticulturalists,
surviving primarily on jelly made from the pith of the sago palm and leafy greens
such as the jointfir spinach (Gnetum gnemon) (Photographs A.4, A.5, A.6, and
A.l). The Au supplement their diet with other vegetables grown in their slash-and-
bum gardens, including taro, sweet potatoes, bananas, pandanus, amaranth, and
papaya; with gathered foods like wild mushrooms, breadfruit, nuts, grubs, and insects
(mainly eaten by children); with animals like snakes, lizards, birds, and bird-eggs
happened upon during daily activities in the bush; more rarely with hunted game
including bandicoot, wild pig, and flying fox; and still more rarely (usually on
prestigious or ceremonial occasions) with domesticated animals, such as pigs and
chickens. Unlike others areas of PNG like the highlands popularized by Rappaport
(1968), pigs are not abundant in Anguganak. In fact, only two domesticated pigs
were observed during the entire field season. Fish are rarely eaten unless they are of
the store-bought tinned variety, as the nearest rivers have few large fish29. Finally, if
they can afford it30, the Au supplement their diet with store-bought rice and instant
noodles, purchased at the trade-store on the mission station (hereafter called the
Station) set up alongside the Anguganak airstrip (Photograph A.8). The two small
(one single-engine and one twin propeller) planes that arrive at the airstrip twice a
week provide the most reliable connection with the nearest port town of Wewak, as
the arduous road between the two is unpaved, often impassable, and sometimes
dangerous due to both road conditions and bandits. The Station is also the location of
29 However, plans are being made in at least one Au village to import tilapia and stock two dug-out
fish ponds that, currently empty, are better classified as mosquito farms.
30 Those who buy food are usually wage-earners employed by either the government, for example, as
nurses or teachers, or by missionaries. Wage-earners are decidedly in the minority, as for example,
only 11.8% of our sample works for wages.
52


the public school that all children have the opportunity to attend if their families can
afford the enrollment fee (Photograph A.9).
Though recent data could not be found, old data on health indicators in
Anguganak are interesting and probably comparable to current numbers31. The infant
mortality rate reported by Tracer (1991) more than a decade ago was 104/1000 live
births, a number unchanged from previous studies up to two decades earlier. The
mean age of marriage for girls was about 21 years, with first birth occurring about 2
years later. The total fertility rate, defined as the mean number of live-births ever
experienced by post-reproductive aged women (over the age of 45), was 6.1. This
number is comparable to other natural fertility populations, or populations that do
not have the intent or the means to control parity (Tracer 1991).
3.2.2 Reciprocal Exchange
Like most Melanesian societies, a complex and rigid system of reciprocal
exchange pervades the social, economic, and political sectors of the lives of the Au,
for example in social grooming debusing (Photograph A. 10), food taboosa
hunter may not consume his kill, but must distribute it among kinand in marriage
practices (Banks 1998, Lewis 1975, Sillitoe 1998, Tracer 2003, see Zimmer-
Tamakoshi 1997 for a detailed account of exchange rules, especially with regard to
land tenure and power relationships). Not only does a mans family pay a brideprice
for his wife, other payments are made to her family at the birth of the first child, the
childs first consumption of meat, puberty, et cetera. In exchange, the womans
brother (the childs maternal uncle) nurtures a special relationship with his niece or
31 Due to the lack of development and therefore lack of medical advancement experienced by the
village, health indicators probably have not changed drastically over the past two decades. In fact,
Anguganak may be less-developed now than it was in the past because of lack of full-time
missionaries and clinic personnel from other countries, and fewer missionary flights to the area with
supplies. The Station was previously home to a post office and a bank, and missionaries installed
amenities such as phones run on electricity from generators. All of these are now gone, save the
generator at the medical clinic. Radio is the only means of communication with Wewak.
53


nephew providing them with protein, performing ceremonial rites, and so on.
Failure to comply with these social norms may result in ostracization or violence.
In addition to formal reciprocal mores, the Au consider it a right to request
everyday items from each other, most frequently betel nut and food, but also more
valuable items like clothing, string bags, tools, and even money (Tracer 2003). If X
requests an item, Y must comply. If Y refuses, she risks being shunned, physically
abused, or at least being ignored should she request an item of someone in the future.
However, should X abuse her rights to request items and do so too frequently, she
may also be shunned or peppered with requests for items. In addition to generosity
with solicited items, the giving of unsolicited gifts, usually of hunted game or other
food, also serves to strengthen social ties (Sahlins 1972, Tracer 2003). Taking a gift
of either kind binds an individual to return the favor at some time in the future, upon
request or otherwise.
While the generosity norm keeps wealth and goods evenly distributed, it also
encourages discreetness with goods and hunted game. Moreover, because the Au
recognize their future obligation when receiving a gift, they sometimes refuse offers
due to unwillingness or inability to pay it back (Tracer 2003). Finally, compensatory
gift-giving after wrongdoing helps to adjust, maintain, restore, redefine, or in the
case of inadequate compensation, break relationships (Banks 1998).
3.2.3 Law and Justice
As a vestige of colonization, the relationship between formal law and custom
has been (and still is) strained in PNG. The imposed western-based, judicial system 32
32
For example, during the field season, a seven-year old child standing under a coconut tree was
struck on the head by a falling coconut and died. Members of the childs village and her mothers
natal village went into mourning, and a series of compensatory exchanges ensued. Most interesting
was the hefty monetary compensation paid by the childs father, his family, and other village members
to the childs maternal uncle, in apology for not taking better care of his niece. The uncle reciprocated
by hosting a feast for the mourners.
54


is largely incongruous with internal belief systems and cultural notions. Formal
codes do not take into account customary beliefs about dispute settlement
especially ideas about swift and violent reprisal for crime, and about victim
compensationinstead imposing Western assumptions about universality, individual
responsibility, guilt, innocence, and the necessity for a protracted trial system.
The enactment of the Criminal Law (Compensation) Act in 1991 was an
attempt to reconcile this incongruity, and to produce a system of law that integrates
both the Western-based criminal justice system and traditional law (Banks 1998). In
accordance with customary practices of victim compensation, the act empowers the
national and district courts to order the guilty to pay compensation in addition to, or
in lieu of, other punitive measures. Still, the geographical isolation of cultural groups
with greatly differing customs, beliefs, and lifestyles, makes a blanket national law
and court system nearly impossible to apply.
The compensation act has been argued to convert violence, injury, or loss into
currency so that wealth items may be accepted as equal return for the wrong (Banks
1998). Though the compensation act is not determinate but procedural, the courts
usually treat the payment of compensation as a mitigating factor in punitive
sentencing. Compensations are determined by the courts according to what each
party (and his family or clan) is willing to give and accept. As qualitative results will
show, compensation in part or in whole makes up what individuals deem as the
appropriate punitive measure for both petty and violent crime.
Compensation comprises two parts, the first being reparation of the crime and
the second a reevaluation of the wronged relationship in a public manner. This
echoes the principle of restorative justice, explored in the next chapter, which not
only requires retribution by the offender but also engages the three actors involved in
a crimethe offender, victim, and the communityin a joint attempt to amend the
wrong. Ideally, the restorative process works to give a voice to actor feelings and
55


issues. This includes public acknowledgment of shame of the offender and
forgiveness by the victim (Cohen 2001).
56


4. Justice
4.1 Types of Justice
4.1.1 Retributive Justice
One modem form of state-controlled reciprocal altruism is the retributive
criminal justice system. Retribution is a manifestation of the old adage an eye for an
eye, but the fact that retribution can refer to either positive or negative returns is
often overlooked in favor of the punitive sense of the word. In the retributive
paradigm of justice, the focus is primarily on the criminal act (Bennett 2002, Daly
2000), secondarily on how that crime represents a violation of the state, and lastly on
the actual harm done to the victim or community (Zehr 1985). Presumably on behalf
of victims but essentially on behalf of wrongs committed to itself, the state takes
action against the criminal and should he be found guilty, determines an appropriate
punishment in order to replace one social injury with another, allegedly to deter
future crime as well as shore up public well-being (Hampton 1984). Punishment is
usually imprisonment (Johnstone 2003) whereby alienation is both a sacrifice and a
symbol of the communitys moral disapproval. In other words, like Foucaults
asylum, the prison is at once a tool of moral uniformity and social ostracization
(Rabinow 1984). Ideally, the imprisoned wrong-doer will proceed through an
emotional journey from guilt to shame (a form of self-imposed punishment) (Bennett
2002).
Imprisonment is based on the reform movement championed by the Quakers
in the early 1800s, wherein criminal behavior was thought to be the result of the
57


criminals corrupt social environment (Griset 1991). To be effective, imprisonment
should rehabilitate, deter, and incapacitate. That is, it aims to produce Foucaultian
docile bodies, stripped of their destructive power by the disciplinarian state and
coerced into the form of a normalized citizen (Foucault 1977). Fundamental to the
effectiveness of this retributive system is stigma that ideally should shame the
perpetrator into reform and discourage others from emulating his criminal behavior.
However, in most societies, stigma lingers long after the actual imprisonment,
effectively ostracizing the criminal and preventing reintegration through post-
imprisonment limits. These might include the inability to obtain a drivers license,
difficulty in getting a job, denial of rights including the right to vote, et cetera
(Johnstone 2003). While effective and re-integrative shaming could prevent
recidivism as well as keep others from imitating criminal acts, stigmatization more
often exiles, humiliates, and may even perpetuate crime (Braithwaite 1996, Cohen
2001, Foucault 1977).
Zehr (1985) argues that throughout history, there has been a dialectic between
two forms of justice: that of the state and that of the community. Whereas state
justice was legal, formal, rational, rigid, and punitive, community justice was
flexible, context-dependent, often negotiated, and frequently restitution-oriented.
Modem (and usually Western) interpretations of justice have taken the state model
wholesale, simultaneously enhancing the central power of the state as the primary
actor and using the prison as its punishment of choice; this is likely no coincidence.
The imbalanced focus on pure punitive measures leaves little or no room for victim
compensation.
Punishment and restoration are inherently different, as punishment is a
determinate means toward reform while restoration is a flexible process with a
potential outcome; punishment is probably not the most effective means to the end of
restoration (Walgrave 2004). Retribution is bom of moral and ethical sentiments
(Hampton 1984), not unlike the principle of fairness. Under conditions of punitive
58


retribution, the offender suffers in what is deemed an equal amount to the victim; he
pays the victim back in suffering so that the amount of suffering is doubled and
spread equally between the two (Walgrave 2004). In the restorative justice
paradigm, the offender actively pays back in reparations, constructively taking
suffering away instead of adding to it (or having it added to on his behalf).
4.1.2 Restorative Justice
Restorative justice is a progressive alternative to retributive justice, bom of
dissatisfaction with exclusionary means of crime control, particularly crowded and
seemingly ineffective prisons (Cohen 2001). In lieu of the punishment-focus of
retributive justice, restorative justice focuses on perpetrator responsibility and
damage reparation after a crime (Johnstone 2003). Control of the reparation process
does not lie solely in the hands of a formal, state-run judicial system, but is also
appropriated in the hands of community members including the victim and
perpetrator of the crime. Ideally, all stakeholdersor everyone affected by the crime
including community memberscollectively decide how best to resolve the harm
done by the crime to victim, offender, and community, and how to prevent
recidivism. State officials and justice agencies act as facilitators of the process. In
this way the common layperson may do justice through a variety of methods. One
of these is the mediation session currently used in Australia and New Zealand for
youth crimes (Daly 2000). The victim and offender communicate directly in a public
setting, and both participate in decision-making, as opposed to sitting inert and
voiceless in a court room while professionals handle all aspects of the transgression
and its reparation (Cohen 2001, Johnstone 2003). Main goals of the restorative
process include 1) healing: of the offender who must seek forgiveness (from victim,
community, and self); of the victim; and of the community at large, whose members
may feel unsafe or betrayed as a result of the crime; however, harm done to the
community is of secondary importance to the direct victim; 2) encouraging the
59


perpetrator to take responsibility for his actions; 3) emotional journey of the
perpetrator from shame to guilt, to regret, and to empathy for the victim and
community, therefore bolstering his tie to the community so that he may be fully
reintegrated; 4) identification of the social psychological issues that all actors have
to confront in order to reclaim the dignity of the wronged and shamed parties so that
all can effectively reintegrate. In this process, remorse of the perpetrator is a key to
restoration, as is forgiveness (Cohen 2001, Zehr 1985). Restorative justice is
processual, not a tangible or fixed outcome.
There is a tendency to associate the restorative process strictly with restitution
made to the victim (Barnett 1977). When found guilty, the perpetrator may have the
option to offer some sort of compensation to the victim, though not necessarily a
monetary remuneration as it is so difficult to attach a dollar sign to trauma
(Johnstone 2003). Other than money, reparation may also include work for the victim
or community (in some form meaningful to the victim) or course attendance
(counseling, anger management, Alcoholics Anonymous, etc.). Of primary
importance is apology, followed by tailored reparation developed from the needs,
both material and emotional, of the parties involved (Marshall 1998). Retributive
compensation, especially when voluntary, is both symbolic and therapeutic: symbolic
as a sign that the perpetrator is accountable for his actions; therapeutic as a potential
alleviator of guilt and shame. Restitution is something the transgressor does, as
opposed to some punishment done to him.
4.1.3 Distributive Justice and Fairness
A third type of justice is distributive justice, a concept innately linked to
fairness (Rawls 1999, Rischer 2002). Rischer (2002) argues that equity is objective,
and therefore not the subjective conjecture of personal tastes. Equity requires that
shares be divided impartially, impersonally, and evenly without bias to any claim or
demand. Contrary to equity, fairness may change over time and place, so that current
60


and prevailing norms dictate its rules (Richer 2002). Though there are cultural and
social rules that produce the guidelines of fairness, fairness is in a large part
individually determined: what is fair to one individual may not be fair to the next.
Thus fairness is not equivalent to distributive equity unless there are equal claims.
Fairness requires thought, intent, and deliberation, so that distributions meet the
needs set out by fairness.
Justice, linked to fairness, is an issue of proportion. Whereas equity would
give every individual an equal share, justice would give each his due. In distributive
justice, equitable distribution is context dependent. Rischer (2002) argues that
fairness belongs injustice, but not in economics, as it is an essentially moral issue.
He cautions that in game theory and economics, the concept of fairness is skewed to
include only the satisfaction of all parties involved. But happiness (satisfaction) is
different than fairness pursued as an instrument of justice. The paramount
consideration for fairness as an aspect of justice is not how an individual fares in
relation to his own claims but how he fares in relation to the rest of the claimants
(p.16). This statement mirrors the Nash equilibrium and principle of relative fitness
maximization. Fairness in the sense of the Nash equilibrium would mean that the
situation is envy-less, or that no one wants any other persons share or thinks that
their share is deficient with regard to anyone elses share.
Though an adequate review of fairness and how individuals define it is
impossible here (see Rawls 1999, Walster et al. 1978), fairness is complicated by
subjective issues such as the incommensurability of different kinds of goods, and the
inability to understand another individuals perspective (historically and
experientially defined). Another curious aspect of fairness is that though individuals
may publicly agree that equity is important and desirable, it is likely that privately,
each individual would prefer more than the publicly-objective fair share (Shroeder et
al. 2003).
61


Finally, fairness is not a concern until competition for a resource is strong
(i.e., when demand is above carrying capacity), because until then, each individual
may take or have as much as he wants (Schroeder et al. 2003). When resources are
limited however, tensions rise with regard to shared distribution, and actions may be
taken to punish those who seem to take more than their share, especially if they seem
to do so with the intent of exploiting others. Echoes of this in game theoretic
situations (Nowak et al. 2000, Rabin 1993) are visible when participants retaliate
against free-riding defectors. In ultimatum games punishers seem to prefer the
equality of a zero-payoff to the inequitable distribution of monies that, though they
produce a non-zero take, favor the exploitative participant. Equity thus seems to
prevail in group dilemmas; but if individuals are indeed rational, decision-making
should be reduced to a cost-benefit analysis of a given situation.
4.2 Justice and Game Theory
In the context of game theory, different forms of justice take on different
connotations than when examined in the context of crime (Schroeder et al. 2003).
Distributive justice is concerned with the differential payoffs participants receive,
thus aligning it with inequality aversion. Retributive justice applies to actions taken
by participants to punish defectors, usually based on an emotional reaction to unfair
behavior, therefore aligning it with Rabins (1993) faimess-of-intent theory.
Restorative justice encompasses measures taken to compensate participants who are
wronged by other participants and represents uncharted territory in experimental
economics. (Re)distribution in a game may improve one individuals position (the re-
distributors) relative to coplayers, or realign all positions through the more fair
distribution of payoffs. Punishment may attempt to shape future behavior of a player
by dissuading him from defecting in the future. But compensation (restorative
justice) in a game will only do restitution for wrongs done by another player. This
highlights a major difference injustice in game contexts: both restorative and
62


retributive paradigms of justice (theoretically) aim at the rehabilitation of a criminal,
but in a game situation, restorative justice only compensates the victim while letting
the transgressor free-ride.
Finally, in social dilemmas and by extension in game situations, individuals
concerned with justice may anticipate not only immediate effects but long-term
impacts (i.e., private and public resources; social dynamics) (Schroeder et al. 2003).
This may be compounded in cultures where anonymous interactions are rare or non-
existent, because participants cannot suddenly ignore enculturated behavior in the
game situation. Thus observed behavior may seem anomalous to economic and
evolutionary predictions, but instead make manifest important cultural and social
norms like those of fairness and justice that operate with a view to future
implications.
The justice system employed by Western nations is on the state-controlled
retributive end of the justice continuum, while the customary system of PNG has
elements of both retribution (usually swiftly employed physical abuse or even death)
and restoration (compensation of the victim and his family). From the above
discussion, it is clear that a retributive, imprisonment-focused justice paradigm
clashes fundamentally with a combination retributiverestorative paradigm, based
on swift punishment and victim compensation. The imposed retributive judicial
system has probably conditioned Papua New Guineans to accept imprisonment as at
least a secondary or supporting form of punishment to more swift punition and
victim compensation, however. This study aims to elucidate whether either or both
retributive or restorative tendencies are visible in economic game behavior.
63


5. Methods
After receiving approval from the Human Subjects Research Committee,
University of Colorado at Denver (#2005-083) (See Appendix C), the data for this
project were collected over a one-month period in June and July 2005 in the three Au
villages of Brugap, Winaluk, and Anguganak. Data collection in each village was
completed in one days time. The villages were within a few hours walking distance
of each other with both Winaluk and Anguganak situated atop the ridgeline. Both
Anguganak and Brugap have populations of about 350 individuals and are
approximately equidistant from the Station and airstrip. The population of Winaluk is
smaller, at approximately 175 individuals, and Winaluk is farther from the Station
than the other two villages so that one must climb up a mountain from the Station
and pass through several hamlets of Anguganak Village in order to reach Winaluk.
5.1 Sample Recruitment
We selected Brugap, Winaluk, and Anguganak after assessing village
willingness to participate in research. We then notified each village via messenger
several days prior to the intended date of data collection. Individuals were told that
they had the opportunity to volunteer for game-like research, and that they would be
paid a nominal show-up fee with the potential to win more money in the research
game.
Our arrival at each village on the morning of data collection was conveyed by
word-of-mouth to any villagers who had chosen to remain in the village on that day
64


(instead of going to the bush to garden and gather food). We asked all interested men
and women, aged 18 and older, to meet us in a central location. As people trickled in,
we announced that in order to participate, each person had to attend an initial
orientation meeting detailing game play. Latecomers would not be allowed to
participate in the action roles of the game, though we did allow some to participate in
the inert roles (see below). While the participants gathered, we recorded the name
and gender of each person in order to record the number of people in attendance, to
attempt to have a gender-balanced group, and to ensure that only those present before
the meeting began were allowed to take on decision-making roles in the game.
5.2 The Game
The third-party justice game is dynamic33 in that there is a sequential order to
player action so that at least some player action is dependent upon previous action by
another player (Romp 1997). It is also a normal form game in that the identity of the
players is anonymous and all players have complete information about game rules.
The original third-party punishment game is a dictator game in that one player
determines how to split a sum with another player, with the addition of a third-party
who is given an endowment equal to 50% of the endowment allotted to the dictator
and the second-party34. The third-party is then given the option to keep his
endowment or punish the dictator, at a cost, should the dictators behavior be deemed
selfish or defective. The current game, devised by Tracer, the principal investigator,
extends the autonomy of the third-party so that not only may he punish the proposer,
he may also compensate the inert second-party, also at an equal cost; or he may both
punish and compensate at double the cost.
13 In static games all players act without knowing what other players will do, as in a sealed bid auction
or the Prisoners Dilemma (Romp 1997).
34 For example, if the dictator is given an endowment of $10 to divide between himself and the
second-party, the third-party is automatically endowed with $5.
65


More explicitly, Player A (the proposer) divides up a sum of 10 kina (K10)35
between himself and an anonymous Player B (the second-party). Player A may offer
any amount to Player B, from K0 to K.10. Player B is inert: B merely receives his
payoff at the end of the game after the third-party (anonymous Player C) has the
opportunity to sanction Player A or compensate Player B.
Player C is given K5 and four options. After hearing Player As offer to
Player B, Player C may
Option 1) Do nothing: he takes his K5 and Player A and Player B receive
payoffs according to Player As proposal.
Option 2) Player C may punish Player A at a cost to himself; that is, he may
pay K1 (20% of his K5 endowment) to take away K3 from Player
A. Both the paid kina (from Player C) and the taken kina (from
Player A) return to the experimenter.
Option 3) Player C may compensate Player B at a cost to himself; that is, he
may pay K1 (20% of his K5 endowment) to add K3 to the
endowment of Player B. Again, the paid kina comes back to the
experimenter, and the experimenter adds K3 to the payoff of Player
B.
Option 4) Player C may do both, that is both punish and compensate at twice
the cost to himself; he may pay K2 (40% of his K5 endowment) to
both take away K3 from Player A and add K3 to the endowment of
Player B. The paid kina goes back to the experimenter, and the
experimenter transfers K3 from the endowment of Player A to that
of Player B.
For example: Player A is presented with 10 kina. Player A offers K3 (30% of K10) to
Player B, intending to keep K.7 (70%). Player C is given K5 and told what Player A
has offered. Player C then has the option to act or walk away.
35 All game transactions were played using the Papua New Guinean currency of kina. At the time of
research, $1 was approximately equal to three kina (K3). The K10 sum used in the game is at the
upper-end of a days wage for an unskilled laborer, and is a large sum especially considering most
village residents have no wage income (Tracer 2003).
66


Option It: Player C does nothing.
Payoffs: Player A=7 Player B=3 Player C=5.
Option 2): Player C pays K1 to take away K3 from Player A.
Payoffs: Player A=4 Player B=3 Player C=4.
Option 3): Player C pays K1 to add K3 to Player B.
Payoffs: Player A=7 Player B=6 Player C=4.
Option 4): Player C pays K2 to do both actions.
Pay offs: Player A=4 Player B=6 Player C=3.
5.3 Data Collection
We explained the conditions of participation and the rules of the game in
Melanesian pidgin, using a script that had been back-translated into English in order
to check for clarity (See Appendix B). We assured individuals of their voluntary
participation; that they would receive K2 for taking part in the research but could
also earn from KO to K10 during the game; that the research seemed like a game but
was really research; and that all game transactions would be anonymous to all other
participants. We warned the individuals that the research process would take several
hours, and that they were disallowed from talking about the game during this time.
Only those aged 18 and over were allowed to participate.
Verbal explanation of the rules of the game was accompanied by cartoon
illustrations, including several examples (See Appendix B). Examples were chosen
specifically to show the wide range of potential offers and therefore actions allowed
in the game. These included offers of K7, K.3, K.5, and the purely selfish offer of KO.
We encouraged individuals to shout out answers to example questions.
After the initial briefing as a group, participants were called up individually,
alternating between males and females when possible. We interviewed each
67


individual in order to collect basic personal information (see Appendix B) and each
individual received his participation fee at the end of the short interview, before
entering the game enclosure (Photograph A. 11). All individuals were designated as
Player A until half of the group had finished the game. We then shuffled the Player A
data sheets containing offers. The remaining individuals were designated as Player C,
acting on randomly matched Player A offers.
Players A and C each entered an enclosed space (a room in the community
school in Brugap, and small meeting houses in Winaluk and Anguganak) to play the
game in the company of the experimenter. On the floor between the participant and
the experimenter were three large pieces of paper bearing cartoons to represent each
player. Ten K1 coins were lined up on the paper for Player A; five K1 coins were
lined up on the paper for Player C (Photograph A. 12). Before the game commenced,
we checked understanding of the rules of the game, and if necessary used examples
for clarification. In both examples and the actual game, the coins were moved to
explicitly demonstrate the payoffs for each player.
To make his proposal offer, Player A moved any amount of coins from zero
to ten onto the paper representing the payoff for Player B. Understanding was
clarified one last time to ensure that Player A realized that under the conditions of his
offer, the payoffs for both Player A and Player B would be as he dictated. However,
Player A was also asked if he realized that Player C had the opportunity to either
punish or compensate by giving up a portion of his endowment.
Before Player C entered the room, coins were moved to represent the
randomly matched proposal offer Player C had the opportunity to act on. We then
clarified that Player C understood the offer and the proposed payoffs to the other
players should Player C choose to abstain from action. In addition, all of Player Cs
optional actions were clarified, manipulating the coins to show potential payoffs.
After Player C made his decision, the coins were again moved to show the final
68


payoffs as a result of his action (or lack thereof). We then asked Player C to explain
why he chose his particular course of action.
Remaining players were assigned to the role of Player B. In the case that
there were not enough individuals to complete all trios, we also allowed some late-
comers who had not been present for the initial orientation session to assume the role
of Player B. Because Player B is inert, it was inconsequential if he had limited or
incomplete understanding of rules of the game and was absent from the orientation
session. Also, due to time constraints and because Player B makes no decisions but
merely receives a payoff, Player B was not always interviewed to collect
demographic variables, though names, gender, and ages were collected. There were
almost always enough latecomers to complete the trios, with most Players B
happening upon the gathering, doing nothing except give their name, gender, and
age, and walking away with a handsome sum. However in a few cases, one person
was assigned the role of Player B for two or more trios. This was usually done when
the Player B payoff for a particular round was K.O.
After all trios had been completed, individuals were called back into the game
enclosure to receive their payoffs. A small number of players were asked to listen to
one or two hypothetical vignettes about a crime and to choose the appropriate action
to amend the crime from a multiple choice bank (See Appendix B).
Measures were taken throughout play to ensure anonymity of player
decisions, of the identity of specific trios, and of the amounts of payoffs. We
distributed payoffs surreptitiously, passing folded bills and coins into the hands of
players during goodbye-hand shakes. However, curiosity proved too much for some
players as well as for some uninvolved observers who furtively attempted to peer
into the room where the game was played, especially during pay-offs.
The sample size from all three villages was 133, or 46 trios with five Player
Bs participating twice (N-\ 33: Player A n =46, Player B n =41, Player C n =46).
Tables 2 and 3 summarize the sample by village, role, and gender. Though the total
69


sample is evenly divided between males and females, when viewed by village there
is an imbalance in the number of males and females especially for Winaluk and
Anguganak; however, this difference is not statistically significant. On the other
hand, even more pronounced is the gender imbalance when viewed by roles. This is
inconsequential for Player B, but for the decision-making roles A and B, women
outnumber men as Player A 3:2, while men outnumber women as Player C by 5:2.
This gender imbalance is statistically significant so that the null hypothesis of no
difference between roles by gender may be rejected (chi-square 12.838, p-value
0.002).
Table 2 Role by Village and Gender Totals
Village Total (N) Player A Player B Player C Males Females
Brugap 63 21 21 21 32 31
Winaluk 29 11 7 11 12 17
Anguganak 41 14 13 14 23 18
TOTALS N=133 46 41 46 67 66
Table 3 Role by Gender and Village
Player A Player B Player C
Village N Male Female Male Female Male Female
Brugap 63 10 11 9 12 13 8
Winaluk 29 2 9 1 6 9 2
Anguganak 41 6 8 6 7 11 3
Totals 133 18 28 16 25 33 13
5.4 Analysis
After entering all data into Microsoft Excel, data were imported into
SPSS 13.0 for Windows for all statistical analyses. For comparison of the means
of groups of nominal data, chi-square tests with Cramers V measures of association
were used. When comparing the means of more than two groups of continuous or
nominal data, one-way ANOVA with post-hoc LSD tests were used. All tests were
70


run with a confidence interval of 95% so that the significance level was flagged for
p-values of <.05. However, exact significance levels are always provided. For
nominal data, cases with missing values were excluded on a test-by-test basis. For
continuous data and logistic regression analyses, cases with missing values were
excluded on a listwise basis.
5.5 Methodological Issues
First, individuals participating as Player A generally had a good grasp of their
task in the game after the initial orientation session, and made their decisions and
offers quickly. Those participating as Player C, however, often required a review
explanation of their role and additional examples after entering the enclosure to play
the game. This is understandable as actions available to Player C are more
complicated. Unfortunately, the difficulty of the Player C task along with
experimenter-bias about cognitive ability, may have contributed to gender imbalance.
Women tend to be less well-educated than men, as a one-way ANOVA shows
statistically (F = 30.036, p-value <0.0001) (See Figures 1 and 2).
Figure 1 Education by Gender
Figure 1: Parentheses indicate percentage of
gender, for example, 88% of all males for whom
data were available reported at least some
education.
Figure 2 Education by Gender and Village
Figure 2: Labels indicate percentage of gender
educated, for example, 100% of males from
Brugap reported at least some education while
only 58% of male participants from Winaluk
reported any education.
71


Participant education is generally predicted by village distance from the Station and
thus the community school. Winaluk was the farthest village from the Station and
had the lowest education rates: 58% of male participants reported at least some
education while only 18% of females reported any education. Our bias may have
predisposed us to designate women for the simpler task of Player A or the inert task
of Player B while reserving those with at least some education for the role of Player
C. Indeed, in Winaluk, where education was much lower than at either Brugap or
Anguganak, frustration with a low level of comprehension definitely drove us to
choose better-educated individuals for difficult roles. The null hypothesis of no
educational difference between villages may be rejected (ANOVA F = 12.464, p-
value <0.0001) while the LSD post hoc test showed that while the mean difference
between Brugap and Anguganak was not significant (0.569, p-value = 0.389), the
mean difference between each of the other villages and Winaluk was significant
(Winaluk and Brugap: -2.993, p-value <0.0001; Winaluk and Anguganak: -3.562, p-
value <0.0001).
As discussed in Chapter 6, there is a secondary modal offer of 0%, contrary
to previous UG experiments among the Au and elsewhere. Such purely selfish offers
seem to corroborate the predictions of economic and evolutionary theory. However,
the way in which the game is framed, which Bolton et al. (1998) assert may bias
game behavior, may have influenced offers. We used carefully chosen language to
indicate that the original endowment of K10 belonged to both Players A and B, but
that allocation was at the discretion of Player A. Though Hoffman et al. (1994) warn
against such a methodology on semantic grounds, citing that this language insinuates
that A must give up some amount of the original bank thus driving offers up, the
extremely high number of K0 offers demonstrates that Players A certainly did not
feel any obligation to make a non-selfish offer. Indeed, to counter this obligatory
sentiment, one of our verbally and pictorially represented examples used in the
orientation session was that of a purely selfish offer of K0. Using this example, and
72


clearly demonstrating the example with cartoons, gives explicit proof that an offer of
0%, even if punished, ensures that Player A leaves with no less than 70% of the
original endowment and that Player B can leave with no more than 30%. The risk of
being selfish is thus very tempting, especially if rational proposers presume the
rationality of other players so that no Player C would altruistically punish or
compensate. It is questionable if the use of an extremely clear example offer of 0%
(among other more altruistic offers) encouraged participants to offer less than they
normally would had they been left to figure out the payoffs on their own, without the
help of tangible, pictorial representations.
Finally, the Player C role is very difficult, requiring considerable explanation.
To prevent wasting time explaining the rules of the game to each player individually,
we chose to explain the game first to all players in a pre-game orientation. Though
they were discouraged from talking about the game, this may have allowed collusion
about game-decisions as players waited several hours for their turn to participate in
the game. This may also explain the high rate of K0 offers, previously absent in other
economic experiments among the Au (Tracer 2003, 2004).
73


6. Results and Discussion
6.1 Quantitative Results
6.1.1 Frequencies of Variables
A total of N = 133 volunteers from the three different villages of Brugap,
Winaluk, and Anguganak participated in the third-party justice game. The sample
included 67 males and 66 females ranging in age from 18 to 80. Table 4 summarizes
the frequencies of scale variables collected during qualitative interviews.
Table 4 Frequencies of Variables
Variable n Minimum Maximum Mode Mean s.d.
Age (yrs) 85 18 80 26 33.13 13.33
# of children 119 0 12 3 2.27 2.48
Education (grade) 118 0 10 0 3.65 3.37
# of gardens* 115 0 11 5 4.63 2.19
Church attendance** 118 0 4 4 2.73 1.69
Cash crop income*** 103 0 1500 100 129.50 240.73
Work income**** 119 0 400 0 15.01 61.30
NOTE: Due to time constraints, all variables were not collected for every participant, particularly for
inert Player B. In addition, some participants either did not understand questions, or did not know the
answer to questions (e.g., did not know their age), so that there are missing data.
* Number ofgaden kaikai, or food gardens.
**Church attendance denotes number of times of attendance per month, assuming that one month has
four Sundays so that the maximum answer is four.
***Cash crop income in Kina per month, usually from selling cocoa, coffee, and/or vanilla.
Wages earned by working in Kina per month. No paying jobs exist in the villages, so that all
participants earn their income at local schools, the clinic (nurses), or at the Station. In fact, of the 115
participants for whom data were available, only 14(11.8%) are wage workers.
Table 5 presents the results of one-way ANOVA comparison of means of all
variables across the three villages. Though several variables differed significantly
74


between villages (See Figures 3 and 4 for distributions of significantly different
means; Figure 3 shows the continuous data associated with the variables of
education, church attendance, number of gardens, and cash crop income, while
Figure 4 shows the nominal data associated with existence of vanilla and cocoa
gardens as a source of cash crop income), ANOVA analyses of Player A offers and
Player C actions show that we cannot reject the null hypothesis of no difference
between villages for game data (Player A offer: F = .439, p-value = .645; Player C
action: F = .446, p-value = .641). Therefore, post hoc multiple comparisons between
the villages are not reported for the categories that were significantly different, and
game data from the three villages are lumped into one sample for further analysis.
Table 5 ANOVA Comparison of Variable Means Between Villages
Variable F p-value
Gender .732 .483
Age (years) .020 .981
Marital Status .432 .650
# of Wives 2.646 .080
# of Kids 1.337 .267
Education 12.464 <.0001
Church Attendance 5.666 .005
# of Gardens 4.238 .017
Coffee* .059 .943
Cacao* 3.123 .048
Vanilla* 3.682 .028
Cash Crop Income/Mo. 9.763 <.0001
Work (yes/no) .116 .891
Work Income/Mo. 2.270 .108
Player A Offer .439 .645
Player C Action .446 .641
Note: Variables with significant differences between villages (p-values < .05) are in bold.
*Coffee, Cacao, and Vanilla are nominal categories wherein participants answered yes or no to
cultivating these cash crops.
75


Figure 3 Distribution of Selected Means Figure 4 Vanilla and Cacao by Village
income.
50
45
40
35
g 30
5 25
? 20
^ 15
10
5
0

(98%X98%)
Jg

I Vanilla
I Kakao
Bnjgap
Anguganak
Figure 4: Labels indicate percentage of
participants who have cash crop income from
either vanilla or cacao; for example, 70% of all
Brugap participants grow and sell at least some
vanilla, while almost all (97%) of participants
from Winaluk have cash crop income from
vanilla.
6.1.2 Player A Offers
Table 6 summarizes the distribution of Player A offers (n = 46) in the third-
party justice game experiment. Offers ranged from KO to K9. Offers produce a multi-
modal distribution, with primary modal offers at K3 and K4, each representing
19.6% of total offers (together making up 39.2% of all offers). Surprisingly, a strong
secondary mode exists at offers of K0, totaling 15.2% of all offers. The mean offer
was K3.30 with a standard deviation of 2.38.
76


Table 6 Frequency of Player A Offer (Kina) at Brugap, Winaluk, and Anguganak
Offer n Percentage
0 7 15.2
1 5 10.9
2 4 8.7
3 9 19.6
4 9 19.6
5 6 13.0
6 1 2.2
7 2 4.3
8 1 2.2
9 2 4.3
10 0 0.0
Total 46 100.0
Mean offer = 3.30, s.d. = 2.375.
Modal offers are in bold. Primary modes = 3 and 4; secondary mode = 0.
The mean offer was 3.33 (s.d. 2.59) for males (n 18, 39.1%) and 3.29 (s.d.
= 2.28) for females (n = 28, 60.9%). A two-tailed t-test reveals that there is no
significant difference between these (t = .066; p-value = .948). Of the other
individual variables, only cash-crop income (Pearson correlation = -.377, p-value =
.013) proved a good predictor of Player A offers. Inversely correlated, the more cash
crop income participants reported, the lower the Player A offer. Regression,
correlation, ANOVA, and univariate analyses show that no other single variable or
combination of variables is a good predictor of Player A offers.
6.1.3 Sanctions and Compensations
Table 7 and Figure 5 show the distributions of Player A offers along with
Player C actions to either punish Player A, compensate Player B, or do both (n = 16
action-takers out of n = 46 Player C). More than a third (34.8%) of all Players C took
some action, all for offers of K0 to K4, so that no Player C took any action for any
77


fair (K5) or better (>K5) offer. Astonishingly, Player C punished and/or compensated
at or above half the time for offers of KO to K3; Player C also punished 22.2% of the
time for offers of K4. In other words, when faced with selfish offers of 0%-30%, the
majority of Players C took action.
Table 7 Frequency of Player A Offer (Kina) and Player C Action at
Brugap,Winaluk, and Anguganak
Offer ft offer ft act ion % Action per Yloffer %of Total Action
0 7 4 57.1 25.0
1 5 3 60.0 18.8
2 4 2 50.0 12.5
3 9 5 55.6 31.2
4 9 2 22.2 12.5
5 6 0
6 1 0
7 2 0
8 1 0
9 2 0
10 0 0
Total 46 16 100.0
Figure 5 Distribution of Player A Offer (Kina) and Player C Action
Player C acted on offers of KO to K3 about half the time and on offers of K4 about a quarter of the
time. No Players C acted on offers of K5 and above.
78


Table 8 and Figure 6 itemize Player C action with offers of KO through K5
while Table 9 shows the distribution of Player C action as a whole and by gender.
Punishments occurred for a wider range of offers (KO to K4) than did compensations
(KO to K3); both occurred over wider ranges than the do both action (KO to K2).
The overall sample is well-balanced for gender (67 males and 66 females);
however, the Player A sub-sample is biased toward a female majority while the
Player C sample is biased toward a heavy male majority: 71.7% of Player C
participants were male, 28.3% were female. Overall, the majority (65.2%) of Players
C chose to do nothing and take the K5 payoff. However, 34.7% of the time Player C
chose to punish, compensate, or do both; more specifically, 30.3% of males and an
incredible 46.2% of females did so. These rates of Player 3 action are extremely
high, especially considering that 26.1% of offers were K5 or above (fair or hyper-
fair). The astonishingly high rate of altruistic punishment and compensation of
anonymous trios stands in stark contrast with economic theory that predicts Player C
should never act and thereby sacrifice a portion of his payoff. As discussed below
however, punishment may be congruent with the evolutionary theory of relative
fitness (or utility) maximization, especially as the tendency to punish may be less a
reaction to fairness and more to do with even payoff distribution. Altruistic
compensation is a bit more difficult to explain theoretically, even if the motivation is
toward even payoff distribution.
Table 8 Player C Action by Offer (truncated at K5 because no action was taken for
offers of K5 or greater)
Offer No Action Punish Compensate Do Both
0 3 1 1 2
1 2 0 2 1
2 2 1 0 1
3 4 2 3 0
4 7 2 0 0
5 6 0 0 0
79


Figure 6 Distribution of Player C Action for Offers 0%-50%
c
o
8
<
>>
o
c
a>
3
O
a>
Do Both
Compensate
Punish
No Action
Offers
Table 9 Player C Decisions Overall and by Gender
(%) n Percentage Male (%) Female
No Action 30 65.2 23 (69.7) 7 (53.8)
Punish 6 13.0 5 (15.2) 1 (7.7)
Compensate 6 13.0 4 (12.1) 2 (15.4)
Do Both 4 8.8 1 (3.0) 3 (23.1)
Total (100.0) 46 100.0 33 (100.0) 13
Notes:
Player C could only punish Player A and only compensate Player B. To punish, Player C gave up K.1
(of his K5 allotment) so that the experimenter would take away K.3 from Player A; to compensate,
Player C gave up K1 so that the experimenter would add K.3 to Player B.
To both punish (Player A) and compensate (Player B), Player C gave up K2 (leaving her with K3) in
order to take away and add K3 to Players A and B, respectively.
Percentages in parentheses are percent within gender.
Of the total Player C sample, males made up 71.7% and females made up 28.3%.
80


Though the sample size was too small to stratify and statistically compare
Player C action by any variable36, it is interesting to note that nearly half of all
women (6 of 13) took action whereas less than a third (10 of 33) of all men chose to
punish, compensate, or do both. Whereas an extraordinary 23.1% of women chose to
sacrifice K2 in order to both punish and compensate, only 3.0% of men (1 of 33)
chose to do so. Also interesting is the fact that the male sample skews toward
punishment while the female contingency skews toward compensation or the
punishment compensation combination. Further research with a well-balanced
gender sample will help elucidate whether gender statistically influences the
tendency to punish, compensate, or do both, and will potentially augment a growing
body of research on gender tendencies in economic games. Previous experimental
results suggest that women are less-selfish than men (Croson and Buchan 1999,
Eckel and Grossman 1998), at least when altruism is expensive (Andreoni and
Vesterlund 2001). Our results similarly suggest that women have a higher tendency
than men to altruistically act in the third-party game, usually to do both, or
compensate, rather than punish. Moreover, ifdoing both may be considered a proxy
of the desire for even distribution, the results may 1) further corroborate previous
research that shows that while men tend to be either perfectly selfish or perfectly
self-less, women tend to share evenly (Andreoni and Vesterlund 2001); and 2) may
prompt debate about theories arguing that punishment is a reaction to unfairness and
means toward justice.
36 When Player C action is divided up into four categories (no action, punish, compensate, punish and
compensate), 75.0% of cells have expected counts less than 5. Even with Player C data collapsed into
two categories (no action and action), 25% of cells still have expected counts less than five.
81


6.2 Qualitative Results
6.2.1 Interviews and Vignettes
Post-game interviews with randomly selected third-parties demonstrate that
players were motivated by a variety of sentiments when deciding to punish,
compensate, do both, or walk away. Of those that walked away, some were apathetic
(let Player A and Player B work it out; its none of my business) while others
were conflicted. The latter wanted to even up the pay-outs between all players, but
were either 1) unwilling to give up money to do so or 2) without recourse, that is,
unable to add or subtract from either players payoff to reach equality (K5:K5:K5)
(for example, when faced with hyper-fair offers since there was no option to reduce
the payoff of Player B). The majority of punishers and those that both punished and
compensated cited the desire to even up distribution of payoffs to equality when
asked about their decisions. The compensators, however, gave more emotional
answers, saying I feel sorry for Player B, or I should help my brother/sister.
Equitable distribution as a motive for third-party action will more fully be discussed
below.
In addition, a very small sub-sample of third-party participants {n = 11.2%)
were asked to listen to one or two vignettes about crime, and describe the appropriate
punishment (See Appendix B for vignettes and See Figure 7). Sixty-three percent
stated that compensation is the appropriate punishment for petty crime (Vignette 1),
at least as a primary punitive measure in conjunction with imprisonment should the
guilty party refuse to pay compensation. Imprisonment alone was satisfactory for
only 9% of the sample, while imprisonment coupled with a beating was the choice of
27% of the sample. Responses to vignettes about more serious and violent crime
varied with respect to imprisonment vs. the death penalty, but always included
compensation to the victims family or clan. Qualitative data thus support the
assertion that, probably as an extension of the reciprocal exchange system that values
82


generosity, compensation as a vehicle for restorative justice is an integral part of
punition.
Figure 7 Vignette Responses
>.
u
c
0
3
IT
0
4.5
4
3.5
3
2.5
2
1.5
1
0.5
0
Compensate Compensate Imprisonment Imprisonment +
Conditional Beating
Appropriate Punitive Measure
Figure 6: Compensate Conditional refers to the responses for which compensation was the first choice
of retribution if the transgressor could be persuaded to do so. If the transgressor refused, either
imprisonment or a beating would be appropriate as a secondary course of punitive action.
6.3 Discussion
Table 10 summarizes offers from previously published ultimatum game
experiments. The range of UG and DG offers from games performed in small-scale
societies (Henrich 2000, Henrich et al. 2001) vary much more than those of UGs
played with university students in a diverse group of industrialized countries
(Pittsburgh; Ljubljana, Slovenia; Jerusalem; and Tokyo) (Roth et al. 1990); and in
Yogyakarta (Indonesia) (Cameron 1995). Whereas the mean student offer was
between 43% and 48% and the mode consistently 50%, mean offers from the small-
scale societies ranged from 26% to 58%, with a modal range of 15% to 50%
(Henrich et al. 2001). Results from previous and current studies in Papua New
83


Guinea are between the student and small-scale society offers. Among the groups
from Papua New Guinea, the mean reported here is lower than previous results at
33%, but the modes are consistent with previous findings.
Table 10 Summary of Mean and Modal Offers in Ultimatum Games
Mean Mode*
University students' 43%-48% 50%
Small-scale societies2 26%-58% 15%-50%
Au (of PNG)3 43% 30%
Gnau (of PNG)3 38% 40%
Current results 33% 30%, 40%
*Hyphens indicate range. Commas separate bimodal offers.
(Roth etal. 1990, Cameron 1995) 2(Henrich et al. 2001) 3(Tracer 2003)
The following are a reiteration of predictions followed by discussion of each.
I. The proposer will never make a non-zero offer.
If individuals are selfish, as economic and evolutionary theory predicts, offers in the
third-party justice game should not exceed 0%, especially under conditions of
anonymity.
Results show primary modes at offers of 30% and 40% (K3 and K4), which is
consistent with offers found in previous ultimatum games performed in the same area
of Papua New Guinea (Tracer 2003). The primary modes are in direct opposition to
Hypothesis I and therefore oppose the predictions of evolutionary and economic
theory. However, results also show a strong secondary mode at perfectly-selfish
offers of K0. Tracer (2003) recorded no offers of 0% in a UG previously played
among the Au.
The high rate of zero offers corroborates Hypothesis I, and therefore seems to
corroborate economic and evolutionary predictions that individuals are selfish. Two
84


considerations may otherwise help explain the high rate of zero offers, one
theoretical and the other methodological. First, should economic theory be correct in
assuming that all players are rational maximizers of utility, these rational individuals
should presume that their peers are likewise rational. Thus, Player A would deduce
that no Player C would punish a low offer because it would decrease Player Cs
utility. Alternatively, this rationality might explain selfish offers in a different way.
In the UG previously performed among the Au by Tracer (2003), the proposer may
reasonably expect punishment from the one he wrongs so that offers are non-selfish.
In the third-party justice game however, the threat of punishment is less severe
because proposer-behavior does not affect third-party payoffs. This lack of perceived
threat could encourage more selfish offers than have previously been observed.
A second methodological possibility as discussed in Chapter 5, is that during
the game orientation session, several verbal and pictorial examples were given to
illustrate potential payoffs according to game decisions37. One of these, among
others, was the offer of K.0. With the cartoon representations, it was very easy to see
that in the worst case scenario of Player C punishing Player A for a purely selfish
offer of zero, Player A would leave the game with no less than 70% of the original
stake (K7). It is possible that even with good understanding of game rules, many
players would not have so clearly envisioned their winnings for a K0 offer without
the pictorial example.
Hyper-low (purely selfish offers of K0) offers stand in stark contrast to
hyper-fair offers, or those greater than 50% of the original sum. 13 .0% of all offers
were hyper-fair (>50%), the existence of which alongside modal offers of 30% and
40% clearly opposes Hypothesis I. However, this finding is consistent with a
previous ultimatum game performed among the Au where 14.5% of offers were
hyper-fair (Tracer 2003). Tracer (2003) argues that the existence (and rejection) of
37 Bolton et al. 1998 emphasize the importance of the game frame on offers.
85


hyper-fair offers in the ultimatum game (rejected >50% of the time) may be
explained by the previously described cultural premium placed on generosity, while
rejections may further be explained by the reluctance to be indebted to anyone that
should give too-substantial a gift. Unfamiliarity with anonymous exchanges also
helps to explain why, even when assured that identities of players would only be
known to the experimenter, both offers and rejection rates offair offers are high.
Even so, extreme, hyper-fair offers of 90% are anomalously over-generous. A
further, unlikely possibility might be experimenter-influenced costly signaling
(Henrich 2000, Hoffman et al. 1994, Hoffman et al. 1996); or that the participants,
who knew the primary investigator very well (after 17-plus years of interaction
during the course of various anthropological studies), wanted to appear non-selfish to
the experimenter for some ulterior motive. This may have been further impacted by
the fact that the primary investigator is known to give villagers his remaining
supplies (kerosene lamps, food, bedding, mosquito nets, flashlights, etc.) at the end
of the field season38. Hoffman et al. 1994 argue that not only may experimenter-
influenced costly signaling drive up Player A offers, it could drive up the rate of
altruistic compensation. Indeed, the rate of Player C action in the current third-party
punishment experiment at 34.7% is slightly higher than the punishment rate of 32.8%
that Tracer (2003) found in an ultimatum game experiment. However, rates of hyper-
fair offers and of third-party action were evenly distributed between villages. For
experimenter-bias to be at work we might have expected that hyper-generosity be
more prevalent in Anguganak, the village where the experimenter resides when he
does his fieldwork; and the village in which most of the post-field-season supplies
are doled out. Nonetheless, high rates of altruistic punishment and compensation, as
well as the existence of non-selfish and hyper-fair offers are in direct opposition to
the economic assumption that individuals are selfish, absolute utility maximizers.
3,1 One participant admitted to the primary investigator that he had lied during his interview about
church attendance in order to look good for the author in pre-game interviews.
86


//. Third-party players will never act because it is costly to do so.
The third-party payoff, 50% of the endowment available for division between the
proposer and inert recipient, is allotted independent of proposer and recipient
behavior. According to both economic and evolutionary theory, the third-party
should never punish or compensate because it costs him 20% to do so without any
potential gain. The third-party should certainly never both punish and compensate
because it costs 40% of his endowment. These assertions and the above hypothesis
are clearly refuted by empirical results. Overall, a non-trivial 34.8 % of third-party
participants punished, compensated, or did both. And when faced with unfair offers
of 0%-30%, the majority of third-parties acted.
Punishment and compensation are equally costly (K1 out of K5, or a 20%
sacrifice). One of the aims of this study is to elucidate whether tendencies based on
retributive or restorative justice play a role in game behavior. In a setting where the
retributive justice system predominates, punishment would be expected to prevail
over compensation. PNG has historically relied on both swift punition (retributive
justice) and victim compensation (restorative justice), so that we might expect to find
a mixture of punishment and compensation in third-party behavior.
Overall, players punished and compensated with equal frequency: 13.0%
punished and 13.0% compensated, so that more than a quarter of third-parties
(26.0%) punished or compensated. Third-parties punished or compensated more
often than they did both actions (8.7% of the time). When viewing third-party
action by gender however, more interesting trends emerge.
Females compensated more than they punished: 15% of females compensated
Player B and 7% of females punished Player A. Astoundingly however, females both
punished and compensated at a greater rate than they did either action; 23% chose to
do both actions, incurring a 40% cost to do so. Males punished more often than they
compensated, and did each action more often than they both punished and
87


Full Text
isolates the Au from the Pacific Ocean. The Au populate lowland tropical rain forest,
typical of an equatorial climate where temperatures vary little annually. Annual
rainfall is very high, exceeding 2.5m. Though the periods from October to March and
from April to September are classified as the wet and dry season respectively, the dry
season is not typified by drought but a reduction in overall rainfall with accompanied
drops in river and spring flow. Malaria is endemic to the area as the climate
encourages the proliferation of anopheline mosquitoes; indeed, malaria is the
preeminent cause of both child and adult mortality. Other health concerns include
dengue fever, tuberculosis, tapeworm, tropical ulcers, and scabies.
3.2 The People and Culture
3.2.1 The Au
Au not only refers to the approximately 10,000 individuals who occupy
about 50 villages in the East Au and West Au census divisions, but is also the name
of the predominant28 language spoken in the area (Tracer 1991). Villages, generally
constructed atop ridgelines, range from less than 100 to more than 500 people. The
spatial layout of the villages, constricted by the width of the ridgeline, usually
consists of several hamlets strung together by mudstone paths. Most likely the result
of missionary influence, the current housing layout and therefore sleeping
arrangement is very different from that described by Lewis (1980) and Tracer (1991).
Mens houses that used to provide sleeping quarters for all men and most boys over
the age of 10 have disappeared. Nuclear families now reside together, some in
ground-level, windowless, thatched-roof houses; but many live in more modem
28 Other languages are spoken within the East and West Au census boundaries, including Gnau (Lewis
1980), Elkei, Ghal, and Yil, which together make up an ancient and unique phylum of languages
ostensibly distinct from other languages outside of the Sepik region, and arguably may provide a link
to the original languages brought to PNG by incipient immigrants arriving from the Malay area
(Tracer 1991).
51


houses built atop stilts, some with screen-covered windows, multiple rooms, and
increasingly, corrugated iron roofs (Photographs A.l, A.2, and A.3).
Tracer (1991,2003) characterizes the Au as forager-horticulturalists,
surviving primarily on jelly made from the pith of the sago palm and leafy greens
such as the jointfir spinach (Gnetum gnemon) (Photographs A.4, A.5, A.6, and
A.l). The Au supplement their diet with other vegetables grown in their slash-and-
bum gardens, including taro, sweet potatoes, bananas, pandanus, amaranth, and
papaya; with gathered foods like wild mushrooms, breadfruit, nuts, grubs, and insects
(mainly eaten by children); with animals like snakes, lizards, birds, and bird-eggs
happened upon during daily activities in the bush; more rarely with hunted game
including bandicoot, wild pig, and flying fox; and still more rarely (usually on
prestigious or ceremonial occasions) with domesticated animals, such as pigs and
chickens. Unlike others areas of PNG like the highlands popularized by Rappaport
(1968), pigs are not abundant in Anguganak. In fact, only two domesticated pigs
were observed during the entire field season. Fish are rarely eaten unless they are of
the store-bought tinned variety, as the nearest rivers have few large fish29. Finally, if
they can afford it30, the Au supplement their diet with store-bought rice and instant
noodles, purchased at the trade-store on the mission station (hereafter called the
Station) set up alongside the Anguganak airstrip (Photograph A.8). The two small
(one single-engine and one twin propeller) planes that arrive at the airstrip twice a
week provide the most reliable connection with the nearest port town of Wewak, as
the arduous road between the two is unpaved, often impassable, and sometimes
dangerous due to both road conditions and bandits. The Station is also the location of
29 However, plans are being made in at least one Au village to import tilapia and stock two dug-out
fish ponds that, currently empty, are better classified as mosquito farms.
30 Those who buy food are usually wage-earners employed by either the government, for example, as
nurses or teachers, or by missionaries. Wage-earners are decidedly in the minority, as for example,
only 11.8% of our sample works for wages.
52


the public school that all children have the opportunity to attend if their families can
afford the enrollment fee (Photograph A.9).
Though recent data could not be found, old data on health indicators in
Anguganak are interesting and probably comparable to current numbers31. The infant
mortality rate reported by Tracer (1991) more than a decade ago was 104/1000 live
births, a number unchanged from previous studies up to two decades earlier. The
mean age of marriage for girls was about 21 years, with first birth occurring about 2
years later. The total fertility rate, defined as the mean number of live-births ever
experienced by post-reproductive aged women (over the age of 45), was 6.1. This
number is comparable to other natural fertility populations, or populations that do
not have the intent or the means to control parity (Tracer 1991).
3.2.2 Reciprocal Exchange
Like most Melanesian societies, a complex and rigid system of reciprocal
exchange pervades the social, economic, and political sectors of the lives of the Au,
for example in social grooming debusing (Photograph A. 10), food taboosa
hunter may not consume his kill, but must distribute it among kinand in marriage
practices (Banks 1998, Lewis 1975, Sillitoe 1998, Tracer 2003, see Zimmer-
Tamakoshi 1997 for a detailed account of exchange rules, especially with regard to
land tenure and power relationships). Not only does a mans family pay a brideprice
for his wife, other payments are made to her family at the birth of the first child, the
childs first consumption of meat, puberty, et cetera. In exchange, the womans
brother (the childs maternal uncle) nurtures a special relationship with his niece or
31 Due to the lack of development and therefore lack of medical advancement experienced by the
village, health indicators probably have not changed drastically over the past two decades. In fact,
Anguganak may be less-developed now than it was in the past because of lack of full-time
missionaries and clinic personnel from other countries, and fewer missionary flights to the area with
supplies. The Station was previously home to a post office and a bank, and missionaries installed
amenities such as phones run on electricity from generators. All of these are now gone, save the
generator at the medical clinic. Radio is the only means of communication with Wewak.
53


nephew providing them with protein, performing ceremonial rites, and so on.
Failure to comply with these social norms may result in ostracization or violence.
In addition to formal reciprocal mores, the Au consider it a right to request
everyday items from each other, most frequently betel nut and food, but also more
valuable items like clothing, string bags, tools, and even money (Tracer 2003). If X
requests an item, Y must comply. If Y refuses, she risks being shunned, physically
abused, or at least being ignored should she request an item of someone in the future.
However, should X abuse her rights to request items and do so too frequently, she
may also be shunned or peppered with requests for items. In addition to generosity
with solicited items, the giving of unsolicited gifts, usually of hunted game or other
food, also serves to strengthen social ties (Sahlins 1972, Tracer 2003). Taking a gift
of either kind binds an individual to return the favor at some time in the future, upon
request or otherwise.
While the generosity norm keeps wealth and goods evenly distributed, it also
encourages discreetness with goods and hunted game. Moreover, because the Au
recognize their future obligation when receiving a gift, they sometimes refuse offers
due to unwillingness or inability to pay it back (Tracer 2003). Finally, compensatory
gift-giving after wrongdoing helps to adjust, maintain, restore, redefine, or in the
case of inadequate compensation, break relationships (Banks 1998).
3.2.3 Law and Justice
As a vestige of colonization, the relationship between formal law and custom
has been (and still is) strained in PNG. The imposed western-based, judicial system 32
32
For example, during the field season, a seven-year old child standing under a coconut tree was
struck on the head by a falling coconut and died. Members of the childs village and her mothers
natal village went into mourning, and a series of compensatory exchanges ensued. Most interesting
was the hefty monetary compensation paid by the childs father, his family, and other village members
to the childs maternal uncle, in apology for not taking better care of his niece. The uncle reciprocated
by hosting a feast for the mourners.
54


is largely incongruous with internal belief systems and cultural notions. Formal
codes do not take into account customary beliefs about dispute settlement
especially ideas about swift and violent reprisal for crime, and about victim
compensationinstead imposing Western assumptions about universality, individual
responsibility, guilt, innocence, and the necessity for a protracted trial system.
The enactment of the Criminal Law (Compensation) Act in 1991 was an
attempt to reconcile this incongruity, and to produce a system of law that integrates
both the Western-based criminal justice system and traditional law (Banks 1998). In
accordance with customary practices of victim compensation, the act empowers the
national and district courts to order the guilty to pay compensation in addition to, or
in lieu of, other punitive measures. Still, the geographical isolation of cultural groups
with greatly differing customs, beliefs, and lifestyles, makes a blanket national law
and court system nearly impossible to apply.
The compensation act has been argued to convert violence, injury, or loss into
currency so that wealth items may be accepted as equal return for the wrong (Banks
1998). Though the compensation act is not determinate but procedural, the courts
usually treat the payment of compensation as a mitigating factor in punitive
sentencing. Compensations are determined by the courts according to what each
party (and his family or clan) is willing to give and accept. As qualitative results will
show, compensation in part or in whole makes up what individuals deem as the
appropriate punitive measure for both petty and violent crime.
Compensation comprises two parts, the first being reparation of the crime and
the second a reevaluation of the wronged relationship in a public manner. This
echoes the principle of restorative justice, explored in the next chapter, which not
only requires retribution by the offender but also engages the three actors involved in
a crimethe offender, victim, and the communityin a joint attempt to amend the
wrong. Ideally, the restorative process works to give a voice to actor feelings and
55


issues. This includes public acknowledgment of shame of the offender and
forgiveness by the victim (Cohen 2001).
56


4. Justice
4.1 Types of Justice
4.1.1 Retributive Justice
One modem form of state-controlled reciprocal altruism is the retributive
criminal justice system. Retribution is a manifestation of the old adage an eye for an
eye, but the fact that retribution can refer to either positive or negative returns is
often overlooked in favor of the punitive sense of the word. In the retributive
paradigm of justice, the focus is primarily on the criminal act (Bennett 2002, Daly
2000), secondarily on how that crime represents a violation of the state, and lastly on
the actual harm done to the victim or community (Zehr 1985). Presumably on behalf
of victims but essentially on behalf of wrongs committed to itself, the state takes
action against the criminal and should he be found guilty, determines an appropriate
punishment in order to replace one social injury with another, allegedly to deter
future crime as well as shore up public well-being (Hampton 1984). Punishment is
usually imprisonment (Johnstone 2003) whereby alienation is both a sacrifice and a
symbol of the communitys moral disapproval. In other words, like Foucaults
asylum, the prison is at once a tool of moral uniformity and social ostracization
(Rabinow 1984). Ideally, the imprisoned wrong-doer will proceed through an
emotional journey from guilt to shame (a form of self-imposed punishment) (Bennett
2002).
Imprisonment is based on the reform movement championed by the Quakers
in the early 1800s, wherein criminal behavior was thought to be the result of the
57


criminals corrupt social environment (Griset 1991). To be effective, imprisonment
should rehabilitate, deter, and incapacitate. That is, it aims to produce Foucaultian
docile bodies, stripped of their destructive power by the disciplinarian state and
coerced into the form of a normalized citizen (Foucault 1977). Fundamental to the
effectiveness of this retributive system is stigma that ideally should shame the
perpetrator into reform and discourage others from emulating his criminal behavior.
However, in most societies, stigma lingers long after the actual imprisonment,
effectively ostracizing the criminal and preventing reintegration through post-
imprisonment limits. These might include the inability to obtain a drivers license,
difficulty in getting a job, denial of rights including the right to vote, et cetera
(Johnstone 2003). While effective and re-integrative shaming could prevent
recidivism as well as keep others from imitating criminal acts, stigmatization more
often exiles, humiliates, and may even perpetuate crime (Braithwaite 1996, Cohen
2001, Foucault 1977).
Zehr (1985) argues that throughout history, there has been a dialectic between
two forms of justice: that of the state and that of the community. Whereas state
justice was legal, formal, rational, rigid, and punitive, community justice was
flexible, context-dependent, often negotiated, and frequently restitution-oriented.
Modem (and usually Western) interpretations of justice have taken the state model
wholesale, simultaneously enhancing the central power of the state as the primary
actor and using the prison as its punishment of choice; this is likely no coincidence.
The imbalanced focus on pure punitive measures leaves little or no room for victim
compensation.
Punishment and restoration are inherently different, as punishment is a
determinate means toward reform while restoration is a flexible process with a
potential outcome; punishment is probably not the most effective means to the end of
restoration (Walgrave 2004). Retribution is bom of moral and ethical sentiments
(Hampton 1984), not unlike the principle of fairness. Under conditions of punitive
58


retribution, the offender suffers in what is deemed an equal amount to the victim; he
pays the victim back in suffering so that the amount of suffering is doubled and
spread equally between the two (Walgrave 2004). In the restorative justice
paradigm, the offender actively pays back in reparations, constructively taking
suffering away instead of adding to it (or having it added to on his behalf).
4.1.2 Restorative Justice
Restorative justice is a progressive alternative to retributive justice, bom of
dissatisfaction with exclusionary means of crime control, particularly crowded and
seemingly ineffective prisons (Cohen 2001). In lieu of the punishment-focus of
retributive justice, restorative justice focuses on perpetrator responsibility and
damage reparation after a crime (Johnstone 2003). Control of the reparation process
does not lie solely in the hands of a formal, state-run judicial system, but is also
appropriated in the hands of community members including the victim and
perpetrator of the crime. Ideally, all stakeholdersor everyone affected by the crime
including community memberscollectively decide how best to resolve the harm
done by the crime to victim, offender, and community, and how to prevent
recidivism. State officials and justice agencies act as facilitators of the process. In
this way the common layperson may do justice through a variety of methods. One
of these is the mediation session currently used in Australia and New Zealand for
youth crimes (Daly 2000). The victim and offender communicate directly in a public
setting, and both participate in decision-making, as opposed to sitting inert and
voiceless in a court room while professionals handle all aspects of the transgression
and its reparation (Cohen 2001, Johnstone 2003). Main goals of the restorative
process include 1) healing: of the offender who must seek forgiveness (from victim,
community, and self); of the victim; and of the community at large, whose members
may feel unsafe or betrayed as a result of the crime; however, harm done to the
community is of secondary importance to the direct victim; 2) encouraging the
59


perpetrator to take responsibility for his actions; 3) emotional journey of the
perpetrator from shame to guilt, to regret, and to empathy for the victim and
community, therefore bolstering his tie to the community so that he may be fully
reintegrated; 4) identification of the social psychological issues that all actors have
to confront in order to reclaim the dignity of the wronged and shamed parties so that
all can effectively reintegrate. In this process, remorse of the perpetrator is a key to
restoration, as is forgiveness (Cohen 2001, Zehr 1985). Restorative justice is
processual, not a tangible or fixed outcome.
There is a tendency to associate the restorative process strictly with restitution
made to the victim (Barnett 1977). When found guilty, the perpetrator may have the
option to offer some sort of compensation to the victim, though not necessarily a
monetary remuneration as it is so difficult to attach a dollar sign to trauma
(Johnstone 2003). Other than money, reparation may also include work for the victim
or community (in some form meaningful to the victim) or course attendance
(counseling, anger management, Alcoholics Anonymous, etc.). Of primary
importance is apology, followed by tailored reparation developed from the needs,
both material and emotional, of the parties involved (Marshall 1998). Retributive
compensation, especially when voluntary, is both symbolic and therapeutic: symbolic
as a sign that the perpetrator is accountable for his actions; therapeutic as a potential
alleviator of guilt and shame. Restitution is something the transgressor does, as
opposed to some punishment done to him.
4.1.3 Distributive Justice and Fairness
A third type of justice is distributive justice, a concept innately linked to
fairness (Rawls 1999, Rischer 2002). Rischer (2002) argues that equity is objective,
and therefore not the subjective conjecture of personal tastes. Equity requires that
shares be divided impartially, impersonally, and evenly without bias to any claim or
demand. Contrary to equity, fairness may change over time and place, so that current
60


and prevailing norms dictate its rules (Richer 2002). Though there are cultural and
social rules that produce the guidelines of fairness, fairness is in a large part
individually determined: what is fair to one individual may not be fair to the next.
Thus fairness is not equivalent to distributive equity unless there are equal claims.
Fairness requires thought, intent, and deliberation, so that distributions meet the
needs set out by fairness.
Justice, linked to fairness, is an issue of proportion. Whereas equity would
give every individual an equal share, justice would give each his due. In distributive
justice, equitable distribution is context dependent. Rischer (2002) argues that
fairness belongs injustice, but not in economics, as it is an essentially moral issue.
He cautions that in game theory and economics, the concept of fairness is skewed to
include only the satisfaction of all parties involved. But happiness (satisfaction) is
different than fairness pursued as an instrument of justice. The paramount
consideration for fairness as an aspect of justice is not how an individual fares in
relation to his own claims but how he fares in relation to the rest of the claimants
(p.16). This statement mirrors the Nash equilibrium and principle of relative fitness
maximization. Fairness in the sense of the Nash equilibrium would mean that the
situation is envy-less, or that no one wants any other persons share or thinks that
their share is deficient with regard to anyone elses share.
Though an adequate review of fairness and how individuals define it is
impossible here (see Rawls 1999, Walster et al. 1978), fairness is complicated by
subjective issues such as the incommensurability of different kinds of goods, and the
inability to understand another individuals perspective (historically and
experientially defined). Another curious aspect of fairness is that though individuals
may publicly agree that equity is important and desirable, it is likely that privately,
each individual would prefer more than the publicly-objective fair share (Shroeder et
al. 2003).
61


Finally, fairness is not a concern until competition for a resource is strong
(i.e., when demand is above carrying capacity), because until then, each individual
may take or have as much as he wants (Schroeder et al. 2003). When resources are
limited however, tensions rise with regard to shared distribution, and actions may be
taken to punish those who seem to take more than their share, especially if they seem
to do so with the intent of exploiting others. Echoes of this in game theoretic
situations (Nowak et al. 2000, Rabin 1993) are visible when participants retaliate
against free-riding defectors. In ultimatum games punishers seem to prefer the
equality of a zero-payoff to the inequitable distribution of monies that, though they
produce a non-zero take, favor the exploitative participant. Equity thus seems to
prevail in group dilemmas; but if individuals are indeed rational, decision-making
should be reduced to a cost-benefit analysis of a given situation.
4.2 Justice and Game Theory
In the context of game theory, different forms of justice take on different
connotations than when examined in the context of crime (Schroeder et al. 2003).
Distributive justice is concerned with the differential payoffs participants receive,
thus aligning it with inequality aversion. Retributive justice applies to actions taken
by participants to punish defectors, usually based on an emotional reaction to unfair
behavior, therefore aligning it with Rabins (1993) faimess-of-intent theory.
Restorative justice encompasses measures taken to compensate participants who are
wronged by other participants and represents uncharted territory in experimental
economics. (Re)distribution in a game may improve one individuals position (the re-
distributors) relative to coplayers, or realign all positions through the more fair
distribution of payoffs. Punishment may attempt to shape future behavior of a player
by dissuading him from defecting in the future. But compensation (restorative
justice) in a game will only do restitution for wrongs done by another player. This
highlights a major difference injustice in game contexts: both restorative and
62


retributive paradigms of justice (theoretically) aim at the rehabilitation of a criminal,
but in a game situation, restorative justice only compensates the victim while letting
the transgressor free-ride.
Finally, in social dilemmas and by extension in game situations, individuals
concerned with justice may anticipate not only immediate effects but long-term
impacts (i.e., private and public resources; social dynamics) (Schroeder et al. 2003).
This may be compounded in cultures where anonymous interactions are rare or non-
existent, because participants cannot suddenly ignore enculturated behavior in the
game situation. Thus observed behavior may seem anomalous to economic and
evolutionary predictions, but instead make manifest important cultural and social
norms like those of fairness and justice that operate with a view to future
implications.
The justice system employed by Western nations is on the state-controlled
retributive end of the justice continuum, while the customary system of PNG has
elements of both retribution (usually swiftly employed physical abuse or even death)
and restoration (compensation of the victim and his family). From the above
discussion, it is clear that a retributive, imprisonment-focused justice paradigm
clashes fundamentally with a combination retributiverestorative paradigm, based
on swift punishment and victim compensation. The imposed retributive judicial
system has probably conditioned Papua New Guineans to accept imprisonment as at
least a secondary or supporting form of punishment to more swift punition and
victim compensation, however. This study aims to elucidate whether either or both
retributive or restorative tendencies are visible in economic game behavior.
63


5. Methods
After receiving approval from the Human Subjects Research Committee,
University of Colorado at Denver (#2005-083) (See Appendix C), the data for this
project were collected over a one-month period in June and July 2005 in the three Au
villages of Brugap, Winaluk, and Anguganak. Data collection in each village was
completed in one days time. The villages were within a few hours walking distance
of each other with both Winaluk and Anguganak situated atop the ridgeline. Both
Anguganak and Brugap have populations of about 350 individuals and are
approximately equidistant from the Station and airstrip. The population of Winaluk is
smaller, at approximately 175 individuals, and Winaluk is farther from the Station
than the other two villages so that one must climb up a mountain from the Station
and pass through several hamlets of Anguganak Village in order to reach Winaluk.
5.1 Sample Recruitment
We selected Brugap, Winaluk, and Anguganak after assessing village
willingness to participate in research. We then notified each village via messenger
several days prior to the intended date of data collection. Individuals were told that
they had the opportunity to volunteer for game-like research, and that they would be
paid a nominal show-up fee with the potential to win more money in the research
game.
Our arrival at each village on the morning of data collection was conveyed by
word-of-mouth to any villagers who had chosen to remain in the village on that day
64


(instead of going to the bush to garden and gather food). We asked all interested men
and women, aged 18 and older, to meet us in a central location. As people trickled in,
we announced that in order to participate, each person had to attend an initial
orientation meeting detailing game play. Latecomers would not be allowed to
participate in the action roles of the game, though we did allow some to participate in
the inert roles (see below). While the participants gathered, we recorded the name
and gender of each person in order to record the number of people in attendance, to
attempt to have a gender-balanced group, and to ensure that only those present before
the meeting began were allowed to take on decision-making roles in the game.
5.2 The Game
The third-party justice game is dynamic33 in that there is a sequential order to
player action so that at least some player action is dependent upon previous action by
another player (Romp 1997). It is also a normal form game in that the identity of the
players is anonymous and all players have complete information about game rules.
The original third-party punishment game is a dictator game in that one player
determines how to split a sum with another player, with the addition of a third-party
who is given an endowment equal to 50% of the endowment allotted to the dictator
and the second-party34. The third-party is then given the option to keep his
endowment or punish the dictator, at a cost, should the dictators behavior be deemed
selfish or defective. The current game, devised by Tracer, the principal investigator,
extends the autonomy of the third-party so that not only may he punish the proposer,
he may also compensate the inert second-party, also at an equal cost; or he may both
punish and compensate at double the cost.
13 In static games all players act without knowing what other players will do, as in a sealed bid auction
or the Prisoners Dilemma (Romp 1997).
34 For example, if the dictator is given an endowment of $10 to divide between himself and the
second-party, the third-party is automatically endowed with $5.
65


More explicitly, Player A (the proposer) divides up a sum of 10 kina (K10)35
between himself and an anonymous Player B (the second-party). Player A may offer
any amount to Player B, from K0 to K.10. Player B is inert: B merely receives his
payoff at the end of the game after the third-party (anonymous Player C) has the
opportunity to sanction Player A or compensate Player B.
Player C is given K5 and four options. After hearing Player As offer to
Player B, Player C may
Option 1) Do nothing: he takes his K5 and Player A and Player B receive
payoffs according to Player As proposal.
Option 2) Player C may punish Player A at a cost to himself; that is, he may
pay K1 (20% of his K5 endowment) to take away K3 from Player
A. Both the paid kina (from Player C) and the taken kina (from
Player A) return to the experimenter.
Option 3) Player C may compensate Player B at a cost to himself; that is, he
may pay K1 (20% of his K5 endowment) to add K3 to the
endowment of Player B. Again, the paid kina comes back to the
experimenter, and the experimenter adds K3 to the payoff of Player
B.
Option 4) Player C may do both, that is both punish and compensate at twice
the cost to himself; he may pay K2 (40% of his K5 endowment) to
both take away K3 from Player A and add K3 to the endowment of
Player B. The paid kina goes back to the experimenter, and the
experimenter transfers K3 from the endowment of Player A to that
of Player B.
For example: Player A is presented with 10 kina. Player A offers K3 (30% of K10) to
Player B, intending to keep K.7 (70%). Player C is given K5 and told what Player A
has offered. Player C then has the option to act or walk away.
35 All game transactions were played using the Papua New Guinean currency of kina. At the time of
research, $1 was approximately equal to three kina (K3). The K10 sum used in the game is at the
upper-end of a days wage for an unskilled laborer, and is a large sum especially considering most
village residents have no wage income (Tracer 2003).
66


Option It: Player C does nothing.
Payoffs: Player A=7 Player B=3 Player C=5.
Option 2): Player C pays K1 to take away K3 from Player A.
Payoffs: Player A=4 Player B=3 Player C=4.
Option 3): Player C pays K1 to add K3 to Player B.
Payoffs: Player A=7 Player B=6 Player C=4.
Option 4): Player C pays K2 to do both actions.
Pay offs: Player A=4 Player B=6 Player C=3.
5.3 Data Collection
We explained the conditions of participation and the rules of the game in
Melanesian pidgin, using a script that had been back-translated into English in order
to check for clarity (See Appendix B). We assured individuals of their voluntary
participation; that they would receive K2 for taking part in the research but could
also earn from KO to K10 during the game; that the research seemed like a game but
was really research; and that all game transactions would be anonymous to all other
participants. We warned the individuals that the research process would take several
hours, and that they were disallowed from talking about the game during this time.
Only those aged 18 and over were allowed to participate.
Verbal explanation of the rules of the game was accompanied by cartoon
illustrations, including several examples (See Appendix B). Examples were chosen
specifically to show the wide range of potential offers and therefore actions allowed
in the game. These included offers of K7, K.3, K.5, and the purely selfish offer of KO.
We encouraged individuals to shout out answers to example questions.
After the initial briefing as a group, participants were called up individually,
alternating between males and females when possible. We interviewed each
67


individual in order to collect basic personal information (see Appendix B) and each
individual received his participation fee at the end of the short interview, before
entering the game enclosure (Photograph A. 11). All individuals were designated as
Player A until half of the group had finished the game. We then shuffled the Player A
data sheets containing offers. The remaining individuals were designated as Player C,
acting on randomly matched Player A offers.
Players A and C each entered an enclosed space (a room in the community
school in Brugap, and small meeting houses in Winaluk and Anguganak) to play the
game in the company of the experimenter. On the floor between the participant and
the experimenter were three large pieces of paper bearing cartoons to represent each
player. Ten K1 coins were lined up on the paper for Player A; five K1 coins were
lined up on the paper for Player C (Photograph A. 12). Before the game commenced,
we checked understanding of the rules of the game, and if necessary used examples
for clarification. In both examples and the actual game, the coins were moved to
explicitly demonstrate the payoffs for each player.
To make his proposal offer, Player A moved any amount of coins from zero
to ten onto the paper representing the payoff for Player B. Understanding was
clarified one last time to ensure that Player A realized that under the conditions of his
offer, the payoffs for both Player A and Player B would be as he dictated. However,
Player A was also asked if he realized that Player C had the opportunity to either
punish or compensate by giving up a portion of his endowment.
Before Player C entered the room, coins were moved to represent the
randomly matched proposal offer Player C had the opportunity to act on. We then
clarified that Player C understood the offer and the proposed payoffs to the other
players should Player C choose to abstain from action. In addition, all of Player Cs
optional actions were clarified, manipulating the coins to show potential payoffs.
After Player C made his decision, the coins were again moved to show the final
68


payoffs as a result of his action (or lack thereof). We then asked Player C to explain
why he chose his particular course of action.
Remaining players were assigned to the role of Player B. In the case that
there were not enough individuals to complete all trios, we also allowed some late-
comers who had not been present for the initial orientation session to assume the role
of Player B. Because Player B is inert, it was inconsequential if he had limited or
incomplete understanding of rules of the game and was absent from the orientation
session. Also, due to time constraints and because Player B makes no decisions but
merely receives a payoff, Player B was not always interviewed to collect
demographic variables, though names, gender, and ages were collected. There were
almost always enough latecomers to complete the trios, with most Players B
happening upon the gathering, doing nothing except give their name, gender, and
age, and walking away with a handsome sum. However in a few cases, one person
was assigned the role of Player B for two or more trios. This was usually done when
the Player B payoff for a particular round was K.O.
After all trios had been completed, individuals were called back into the game
enclosure to receive their payoffs. A small number of players were asked to listen to
one or two hypothetical vignettes about a crime and to choose the appropriate action
to amend the crime from a multiple choice bank (See Appendix B).
Measures were taken throughout play to ensure anonymity of player
decisions, of the identity of specific trios, and of the amounts of payoffs. We
distributed payoffs surreptitiously, passing folded bills and coins into the hands of
players during goodbye-hand shakes. However, curiosity proved too much for some
players as well as for some uninvolved observers who furtively attempted to peer
into the room where the game was played, especially during pay-offs.
The sample size from all three villages was 133, or 46 trios with five Player
Bs participating twice (N-\ 33: Player A n =46, Player B n =41, Player C n =46).
Tables 2 and 3 summarize the sample by village, role, and gender. Though the total
69


sample is evenly divided between males and females, when viewed by village there
is an imbalance in the number of males and females especially for Winaluk and
Anguganak; however, this difference is not statistically significant. On the other
hand, even more pronounced is the gender imbalance when viewed by roles. This is
inconsequential for Player B, but for the decision-making roles A and B, women
outnumber men as Player A 3:2, while men outnumber women as Player C by 5:2.
This gender imbalance is statistically significant so that the null hypothesis of no
difference between roles by gender may be rejected (chi-square 12.838, p-value
0.002).
Table 2 Role by Village and Gender Totals
Village Total (N) Player A Player B Player C Males Females
Brugap 63 21 21 21 32 31
Winaluk 29 11 7 11 12 17
Anguganak 41 14 13 14 23 18
TOTALS N=133 46 41 46 67 66
Table 3 Role by Gender and Village
Player A Player B Player C
Village N Male Female Male Female Male Female
Brugap 63 10 11 9 12 13 8
Winaluk 29 2 9 1 6 9 2
Anguganak 41 6 8 6 7 11 3
Totals 133 18 28 16 25 33 13
5.4 Analysis
After entering all data into Microsoft Excel, data were imported into
SPSS 13.0 for Windows for all statistical analyses. For comparison of the means
of groups of nominal data, chi-square tests with Cramers V measures of association
were used. When comparing the means of more than two groups of continuous or
nominal data, one-way ANOVA with post-hoc LSD tests were used. All tests were
70


run with a confidence interval of 95% so that the significance level was flagged for
p-values of <.05. However, exact significance levels are always provided. For
nominal data, cases with missing values were excluded on a test-by-test basis. For
continuous data and logistic regression analyses, cases with missing values were
excluded on a listwise basis.
5.5 Methodological Issues
First, individuals participating as Player A generally had a good grasp of their
task in the game after the initial orientation session, and made their decisions and
offers quickly. Those participating as Player C, however, often required a review
explanation of their role and additional examples after entering the enclosure to play
the game. This is understandable as actions available to Player C are more
complicated. Unfortunately, the difficulty of the Player C task along with
experimenter-bias about cognitive ability, may have contributed to gender imbalance.
Women tend to be less well-educated than men, as a one-way ANOVA shows
statistically (F = 30.036, p-value <0.0001) (See Figures 1 and 2).
Figure 1 Education by Gender
Figure 1: Parentheses indicate percentage of
gender, for example, 88% of all males for whom
data were available reported at least some
education.
Figure 2 Education by Gender and Village
Figure 2: Labels indicate percentage of gender
educated, for example, 100% of males from
Brugap reported at least some education while
only 58% of male participants from Winaluk
reported any education.
71


Participant education is generally predicted by village distance from the Station and
thus the community school. Winaluk was the farthest village from the Station and
had the lowest education rates: 58% of male participants reported at least some
education while only 18% of females reported any education. Our bias may have
predisposed us to designate women for the simpler task of Player A or the inert task
of Player B while reserving those with at least some education for the role of Player
C. Indeed, in Winaluk, where education was much lower than at either Brugap or
Anguganak, frustration with a low level of comprehension definitely drove us to
choose better-educated individuals for difficult roles. The null hypothesis of no
educational difference between villages may be rejected (ANOVA F = 12.464, p-
value <0.0001) while the LSD post hoc test showed that while the mean difference
between Brugap and Anguganak was not significant (0.569, p-value = 0.389), the
mean difference between each of the other villages and Winaluk was significant
(Winaluk and Brugap: -2.993, p-value <0.0001; Winaluk and Anguganak: -3.562, p-
value <0.0001).
As discussed in Chapter 6, there is a secondary modal offer of 0%, contrary
to previous UG experiments among the Au and elsewhere. Such purely selfish offers
seem to corroborate the predictions of economic and evolutionary theory. However,
the way in which the game is framed, which Bolton et al. (1998) assert may bias
game behavior, may have influenced offers. We used carefully chosen language to
indicate that the original endowment of K10 belonged to both Players A and B, but
that allocation was at the discretion of Player A. Though Hoffman et al. (1994) warn
against such a methodology on semantic grounds, citing that this language insinuates
that A must give up some amount of the original bank thus driving offers up, the
extremely high number of K0 offers demonstrates that Players A certainly did not
feel any obligation to make a non-selfish offer. Indeed, to counter this obligatory
sentiment, one of our verbally and pictorially represented examples used in the
orientation session was that of a purely selfish offer of K0. Using this example, and
72


clearly demonstrating the example with cartoons, gives explicit proof that an offer of
0%, even if punished, ensures that Player A leaves with no less than 70% of the
original endowment and that Player B can leave with no more than 30%. The risk of
being selfish is thus very tempting, especially if rational proposers presume the
rationality of other players so that no Player C would altruistically punish or
compensate. It is questionable if the use of an extremely clear example offer of 0%
(among other more altruistic offers) encouraged participants to offer less than they
normally would had they been left to figure out the payoffs on their own, without the
help of tangible, pictorial representations.
Finally, the Player C role is very difficult, requiring considerable explanation.
To prevent wasting time explaining the rules of the game to each player individually,
we chose to explain the game first to all players in a pre-game orientation. Though
they were discouraged from talking about the game, this may have allowed collusion
about game-decisions as players waited several hours for their turn to participate in
the game. This may also explain the high rate of K0 offers, previously absent in other
economic experiments among the Au (Tracer 2003, 2004).
73


6. Results and Discussion
6.1 Quantitative Results
6.1.1 Frequencies of Variables
A total of N = 133 volunteers from the three different villages of Brugap,
Winaluk, and Anguganak participated in the third-party justice game. The sample
included 67 males and 66 females ranging in age from 18 to 80. Table 4 summarizes
the frequencies of scale variables collected during qualitative interviews.
Table 4 Frequencies of Variables
Variable n Minimum Maximum Mode Mean s.d.
Age (yrs) 85 18 80 26 33.13 13.33
# of children 119 0 12 3 2.27 2.48
Education (grade) 118 0 10 0 3.65 3.37
# of gardens* 115 0 11 5 4.63 2.19
Church attendance** 118 0 4 4 2.73 1.69
Cash crop income*** 103 0 1500 100 129.50 240.73
Work income**** 119 0 400 0 15.01 61.30
NOTE: Due to time constraints, all variables were not collected for every participant, particularly for
inert Player B. In addition, some participants either did not understand questions, or did not know the
answer to questions (e.g., did not know their age), so that there are missing data.
* Number ofgaden kaikai, or food gardens.
**Church attendance denotes number of times of attendance per month, assuming that one month has
four Sundays so that the maximum answer is four.
***Cash crop income in Kina per month, usually from selling cocoa, coffee, and/or vanilla.
Wages earned by working in Kina per month. No paying jobs exist in the villages, so that all
participants earn their income at local schools, the clinic (nurses), or at the Station. In fact, of the 115
participants for whom data were available, only 14(11.8%) are wage workers.
Table 5 presents the results of one-way ANOVA comparison of means of all
variables across the three villages. Though several variables differed significantly
74


between villages (See Figures 3 and 4 for distributions of significantly different
means; Figure 3 shows the continuous data associated with the variables of
education, church attendance, number of gardens, and cash crop income, while
Figure 4 shows the nominal data associated with existence of vanilla and cocoa
gardens as a source of cash crop income), ANOVA analyses of Player A offers and
Player C actions show that we cannot reject the null hypothesis of no difference
between villages for game data (Player A offer: F = .439, p-value = .645; Player C
action: F = .446, p-value = .641). Therefore, post hoc multiple comparisons between
the villages are not reported for the categories that were significantly different, and
game data from the three villages are lumped into one sample for further analysis.
Table 5 ANOVA Comparison of Variable Means Between Villages
Variable F p-value
Gender .732 .483
Age (years) .020 .981
Marital Status .432 .650
# of Wives 2.646 .080
# of Kids 1.337 .267
Education 12.464 <.0001
Church Attendance 5.666 .005
# of Gardens 4.238 .017
Coffee* .059 .943
Cacao* 3.123 .048
Vanilla* 3.682 .028
Cash Crop Income/Mo. 9.763 <.0001
Work (yes/no) .116 .891
Work Income/Mo. 2.270 .108
Player A Offer .439 .645
Player C Action .446 .641
Note: Variables with significant differences between villages (p-values < .05) are in bold.
*Coffee, Cacao, and Vanilla are nominal categories wherein participants answered yes or no to
cultivating these cash crops.
75


Figure 3 Distribution of Selected Means Figure 4 Vanilla and Cacao by Village
income.
50
45
40
35
g 30
5 25
? 20
^ 15
10
5
0

(98%X98%)
Jg

I Vanilla
I Kakao
Bnjgap
Anguganak
Figure 4: Labels indicate percentage of
participants who have cash crop income from
either vanilla or cacao; for example, 70% of all
Brugap participants grow and sell at least some
vanilla, while almost all (97%) of participants
from Winaluk have cash crop income from
vanilla.
6.1.2 Player A Offers
Table 6 summarizes the distribution of Player A offers (n = 46) in the third-
party justice game experiment. Offers ranged from KO to K9. Offers produce a multi-
modal distribution, with primary modal offers at K3 and K4, each representing
19.6% of total offers (together making up 39.2% of all offers). Surprisingly, a strong
secondary mode exists at offers of K0, totaling 15.2% of all offers. The mean offer
was K3.30 with a standard deviation of 2.38.
76


Table 6 Frequency of Player A Offer (Kina) at Brugap, Winaluk, and Anguganak
Offer n Percentage
0 7 15.2
1 5 10.9
2 4 8.7
3 9 19.6
4 9 19.6
5 6 13.0
6 1 2.2
7 2 4.3
8 1 2.2
9 2 4.3
10 0 0.0
Total 46 100.0
Mean offer = 3.30, s.d. = 2.375.
Modal offers are in bold. Primary modes = 3 and 4; secondary mode = 0.
The mean offer was 3.33 (s.d. 2.59) for males (n 18, 39.1%) and 3.29 (s.d.
= 2.28) for females (n = 28, 60.9%). A two-tailed t-test reveals that there is no
significant difference between these (t = .066; p-value = .948). Of the other
individual variables, only cash-crop income (Pearson correlation = -.377, p-value =
.013) proved a good predictor of Player A offers. Inversely correlated, the more cash
crop income participants reported, the lower the Player A offer. Regression,
correlation, ANOVA, and univariate analyses show that no other single variable or
combination of variables is a good predictor of Player A offers.
6.1.3 Sanctions and Compensations
Table 7 and Figure 5 show the distributions of Player A offers along with
Player C actions to either punish Player A, compensate Player B, or do both (n = 16
action-takers out of n = 46 Player C). More than a third (34.8%) of all Players C took
some action, all for offers of K0 to K4, so that no Player C took any action for any
77


fair (K5) or better (>K5) offer. Astonishingly, Player C punished and/or compensated
at or above half the time for offers of KO to K3; Player C also punished 22.2% of the
time for offers of K4. In other words, when faced with selfish offers of 0%-30%, the
majority of Players C took action.
Table 7 Frequency of Player A Offer (Kina) and Player C Action at
Brugap,Winaluk, and Anguganak
Offer ft offer ft act ion % Action per Yloffer %of Total Action
0 7 4 57.1 25.0
1 5 3 60.0 18.8
2 4 2 50.0 12.5
3 9 5 55.6 31.2
4 9 2 22.2 12.5
5 6 0
6 1 0
7 2 0
8 1 0
9 2 0
10 0 0
Total 46 16 100.0
Figure 5 Distribution of Player A Offer (Kina) and Player C Action
Player C acted on offers of KO to K3 about half the time and on offers of K4 about a quarter of the
time. No Players C acted on offers of K5 and above.
78


Table 8 and Figure 6 itemize Player C action with offers of KO through K5
while Table 9 shows the distribution of Player C action as a whole and by gender.
Punishments occurred for a wider range of offers (KO to K4) than did compensations
(KO to K3); both occurred over wider ranges than the do both action (KO to K2).
The overall sample is well-balanced for gender (67 males and 66 females);
however, the Player A sub-sample is biased toward a female majority while the
Player C sample is biased toward a heavy male majority: 71.7% of Player C
participants were male, 28.3% were female. Overall, the majority (65.2%) of Players
C chose to do nothing and take the K5 payoff. However, 34.7% of the time Player C
chose to punish, compensate, or do both; more specifically, 30.3% of males and an
incredible 46.2% of females did so. These rates of Player 3 action are extremely
high, especially considering that 26.1% of offers were K5 or above (fair or hyper-
fair). The astonishingly high rate of altruistic punishment and compensation of
anonymous trios stands in stark contrast with economic theory that predicts Player C
should never act and thereby sacrifice a portion of his payoff. As discussed below
however, punishment may be congruent with the evolutionary theory of relative
fitness (or utility) maximization, especially as the tendency to punish may be less a
reaction to fairness and more to do with even payoff distribution. Altruistic
compensation is a bit more difficult to explain theoretically, even if the motivation is
toward even payoff distribution.
Table 8 Player C Action by Offer (truncated at K5 because no action was taken for
offers of K5 or greater)
Offer No Action Punish Compensate Do Both
0 3 1 1 2
1 2 0 2 1
2 2 1 0 1
3 4 2 3 0
4 7 2 0 0
5 6 0 0 0
79


Figure 6 Distribution of Player C Action for Offers 0%-50%
c
o
8
<
>>
o
c
a>
3
O
a>
Do Both
Compensate
Punish
No Action
Offers
Table 9 Player C Decisions Overall and by Gender
(%) n Percentage Male (%) Female
No Action 30 65.2 23 (69.7) 7 (53.8)
Punish 6 13.0 5 (15.2) 1 (7.7)
Compensate 6 13.0 4 (12.1) 2 (15.4)
Do Both 4 8.8 1 (3.0) 3 (23.1)
Total (100.0) 46 100.0 33 (100.0) 13
Notes:
Player C could only punish Player A and only compensate Player B. To punish, Player C gave up K.1
(of his K5 allotment) so that the experimenter would take away K.3 from Player A; to compensate,
Player C gave up K1 so that the experimenter would add K.3 to Player B.
To both punish (Player A) and compensate (Player B), Player C gave up K2 (leaving her with K3) in
order to take away and add K3 to Players A and B, respectively.
Percentages in parentheses are percent within gender.
Of the total Player C sample, males made up 71.7% and females made up 28.3%.
80


Though the sample size was too small to stratify and statistically compare
Player C action by any variable36, it is interesting to note that nearly half of all
women (6 of 13) took action whereas less than a third (10 of 33) of all men chose to
punish, compensate, or do both. Whereas an extraordinary 23.1% of women chose to
sacrifice K2 in order to both punish and compensate, only 3.0% of men (1 of 33)
chose to do so. Also interesting is the fact that the male sample skews toward
punishment while the female contingency skews toward compensation or the
punishment compensation combination. Further research with a well-balanced
gender sample will help elucidate whether gender statistically influences the
tendency to punish, compensate, or do both, and will potentially augment a growing
body of research on gender tendencies in economic games. Previous experimental
results suggest that women are less-selfish than men (Croson and Buchan 1999,
Eckel and Grossman 1998), at least when altruism is expensive (Andreoni and
Vesterlund 2001). Our results similarly suggest that women have a higher tendency
than men to altruistically act in the third-party game, usually to do both, or
compensate, rather than punish. Moreover, ifdoing both may be considered a proxy
of the desire for even distribution, the results may 1) further corroborate previous
research that shows that while men tend to be either perfectly selfish or perfectly
self-less, women tend to share evenly (Andreoni and Vesterlund 2001); and 2) may
prompt debate about theories arguing that punishment is a reaction to unfairness and
means toward justice.
36 When Player C action is divided up into four categories (no action, punish, compensate, punish and
compensate), 75.0% of cells have expected counts less than 5. Even with Player C data collapsed into
two categories (no action and action), 25% of cells still have expected counts less than five.
81


6.2 Qualitative Results
6.2.1 Interviews and Vignettes
Post-game interviews with randomly selected third-parties demonstrate that
players were motivated by a variety of sentiments when deciding to punish,
compensate, do both, or walk away. Of those that walked away, some were apathetic
(let Player A and Player B work it out; its none of my business) while others
were conflicted. The latter wanted to even up the pay-outs between all players, but
were either 1) unwilling to give up money to do so or 2) without recourse, that is,
unable to add or subtract from either players payoff to reach equality (K5:K5:K5)
(for example, when faced with hyper-fair offers since there was no option to reduce
the payoff of Player B). The majority of punishers and those that both punished and
compensated cited the desire to even up distribution of payoffs to equality when
asked about their decisions. The compensators, however, gave more emotional
answers, saying I feel sorry for Player B, or I should help my brother/sister.
Equitable distribution as a motive for third-party action will more fully be discussed
below.
In addition, a very small sub-sample of third-party participants {n = 11.2%)
were asked to listen to one or two vignettes about crime, and describe the appropriate
punishment (See Appendix B for vignettes and See Figure 7). Sixty-three percent
stated that compensation is the appropriate punishment for petty crime (Vignette 1),
at least as a primary punitive measure in conjunction with imprisonment should the
guilty party refuse to pay compensation. Imprisonment alone was satisfactory for
only 9% of the sample, while imprisonment coupled with a beating was the choice of
27% of the sample. Responses to vignettes about more serious and violent crime
varied with respect to imprisonment vs. the death penalty, but always included
compensation to the victims family or clan. Qualitative data thus support the
assertion that, probably as an extension of the reciprocal exchange system that values
82


generosity, compensation as a vehicle for restorative justice is an integral part of
punition.
Figure 7 Vignette Responses
>.
u
c
0
3
IT
0
4.5
4
3.5
3
2.5
2
1.5
1
0.5
0
Compensate Compensate Imprisonment Imprisonment +
Conditional Beating
Appropriate Punitive Measure
Figure 6: Compensate Conditional refers to the responses for which compensation was the first choice
of retribution if the transgressor could be persuaded to do so. If the transgressor refused, either
imprisonment or a beating would be appropriate as a secondary course of punitive action.
6.3 Discussion
Table 10 summarizes offers from previously published ultimatum game
experiments. The range of UG and DG offers from games performed in small-scale
societies (Henrich 2000, Henrich et al. 2001) vary much more than those of UGs
played with university students in a diverse group of industrialized countries
(Pittsburgh; Ljubljana, Slovenia; Jerusalem; and Tokyo) (Roth et al. 1990); and in
Yogyakarta (Indonesia) (Cameron 1995). Whereas the mean student offer was
between 43% and 48% and the mode consistently 50%, mean offers from the small-
scale societies ranged from 26% to 58%, with a modal range of 15% to 50%
(Henrich et al. 2001). Results from previous and current studies in Papua New
83


Guinea are between the student and small-scale society offers. Among the groups
from Papua New Guinea, the mean reported here is lower than previous results at
33%, but the modes are consistent with previous findings.
Table 10 Summary of Mean and Modal Offers in Ultimatum Games
Mean Mode*
University students' 43%-48% 50%
Small-scale societies2 26%-58% 15%-50%
Au (of PNG)3 43% 30%
Gnau (of PNG)3 38% 40%
Current results 33% 30%, 40%
*Hyphens indicate range. Commas separate bimodal offers.
(Roth etal. 1990, Cameron 1995) 2(Henrich et al. 2001) 3(Tracer 2003)
The following are a reiteration of predictions followed by discussion of each.
I. The proposer will never make a non-zero offer.
If individuals are selfish, as economic and evolutionary theory predicts, offers in the
third-party justice game should not exceed 0%, especially under conditions of
anonymity.
Results show primary modes at offers of 30% and 40% (K3 and K4), which is
consistent with offers found in previous ultimatum games performed in the same area
of Papua New Guinea (Tracer 2003). The primary modes are in direct opposition to
Hypothesis I and therefore oppose the predictions of evolutionary and economic
theory. However, results also show a strong secondary mode at perfectly-selfish
offers of K0. Tracer (2003) recorded no offers of 0% in a UG previously played
among the Au.
The high rate of zero offers corroborates Hypothesis I, and therefore seems to
corroborate economic and evolutionary predictions that individuals are selfish. Two
84


considerations may otherwise help explain the high rate of zero offers, one
theoretical and the other methodological. First, should economic theory be correct in
assuming that all players are rational maximizers of utility, these rational individuals
should presume that their peers are likewise rational. Thus, Player A would deduce
that no Player C would punish a low offer because it would decrease Player Cs
utility. Alternatively, this rationality might explain selfish offers in a different way.
In the UG previously performed among the Au by Tracer (2003), the proposer may
reasonably expect punishment from the one he wrongs so that offers are non-selfish.
In the third-party justice game however, the threat of punishment is less severe
because proposer-behavior does not affect third-party payoffs. This lack of perceived
threat could encourage more selfish offers than have previously been observed.
A second methodological possibility as discussed in Chapter 5, is that during
the game orientation session, several verbal and pictorial examples were given to
illustrate potential payoffs according to game decisions37. One of these, among
others, was the offer of K.0. With the cartoon representations, it was very easy to see
that in the worst case scenario of Player C punishing Player A for a purely selfish
offer of zero, Player A would leave the game with no less than 70% of the original
stake (K7). It is possible that even with good understanding of game rules, many
players would not have so clearly envisioned their winnings for a K0 offer without
the pictorial example.
Hyper-low (purely selfish offers of K0) offers stand in stark contrast to
hyper-fair offers, or those greater than 50% of the original sum. 13 .0% of all offers
were hyper-fair (>50%), the existence of which alongside modal offers of 30% and
40% clearly opposes Hypothesis I. However, this finding is consistent with a
previous ultimatum game performed among the Au where 14.5% of offers were
hyper-fair (Tracer 2003). Tracer (2003) argues that the existence (and rejection) of
37 Bolton et al. 1998 emphasize the importance of the game frame on offers.
85


hyper-fair offers in the ultimatum game (rejected >50% of the time) may be
explained by the previously described cultural premium placed on generosity, while
rejections may further be explained by the reluctance to be indebted to anyone that
should give too-substantial a gift. Unfamiliarity with anonymous exchanges also
helps to explain why, even when assured that identities of players would only be
known to the experimenter, both offers and rejection rates offair offers are high.
Even so, extreme, hyper-fair offers of 90% are anomalously over-generous. A
further, unlikely possibility might be experimenter-influenced costly signaling
(Henrich 2000, Hoffman et al. 1994, Hoffman et al. 1996); or that the participants,
who knew the primary investigator very well (after 17-plus years of interaction
during the course of various anthropological studies), wanted to appear non-selfish to
the experimenter for some ulterior motive. This may have been further impacted by
the fact that the primary investigator is known to give villagers his remaining
supplies (kerosene lamps, food, bedding, mosquito nets, flashlights, etc.) at the end
of the field season38. Hoffman et al. 1994 argue that not only may experimenter-
influenced costly signaling drive up Player A offers, it could drive up the rate of
altruistic compensation. Indeed, the rate of Player C action in the current third-party
punishment experiment at 34.7% is slightly higher than the punishment rate of 32.8%
that Tracer (2003) found in an ultimatum game experiment. However, rates of hyper-
fair offers and of third-party action were evenly distributed between villages. For
experimenter-bias to be at work we might have expected that hyper-generosity be
more prevalent in Anguganak, the village where the experimenter resides when he
does his fieldwork; and the village in which most of the post-field-season supplies
are doled out. Nonetheless, high rates of altruistic punishment and compensation, as
well as the existence of non-selfish and hyper-fair offers are in direct opposition to
the economic assumption that individuals are selfish, absolute utility maximizers.
3,1 One participant admitted to the primary investigator that he had lied during his interview about
church attendance in order to look good for the author in pre-game interviews.
86


//. Third-party players will never act because it is costly to do so.
The third-party payoff, 50% of the endowment available for division between the
proposer and inert recipient, is allotted independent of proposer and recipient
behavior. According to both economic and evolutionary theory, the third-party
should never punish or compensate because it costs him 20% to do so without any
potential gain. The third-party should certainly never both punish and compensate
because it costs 40% of his endowment. These assertions and the above hypothesis
are clearly refuted by empirical results. Overall, a non-trivial 34.8 % of third-party
participants punished, compensated, or did both. And when faced with unfair offers
of 0%-30%, the majority of third-parties acted.
Punishment and compensation are equally costly (K1 out of K5, or a 20%
sacrifice). One of the aims of this study is to elucidate whether tendencies based on
retributive or restorative justice play a role in game behavior. In a setting where the
retributive justice system predominates, punishment would be expected to prevail
over compensation. PNG has historically relied on both swift punition (retributive
justice) and victim compensation (restorative justice), so that we might expect to find
a mixture of punishment and compensation in third-party behavior.
Overall, players punished and compensated with equal frequency: 13.0%
punished and 13.0% compensated, so that more than a quarter of third-parties
(26.0%) punished or compensated. Third-parties punished or compensated more
often than they did both actions (8.7% of the time). When viewing third-party
action by gender however, more interesting trends emerge.
Females compensated more than they punished: 15% of females compensated
Player B and 7% of females punished Player A. Astoundingly however, females both
punished and compensated at a greater rate than they did either action; 23% chose to
do both actions, incurring a 40% cost to do so. Males punished more often than they
compensated, and did each action more often than they both punished and
87


compensated: 15% of males punished, 12% compensated, and 3% did both actions.
Though the sample size was not large enough to show a statistical difference in
gender-related behavior, further study with a larger and gender-balanced sample will
help determine if gender indeed is statistically correlated with third-party behavior
and the tendency to punish or compensate.
Altruistic punishment in ultimatum gamesrejecting non-zero offers
conflicts with canonical evolutionary and economic theory and has been argued to be
an emotionally driven reaction to unfair (or selfish) behavior (Bolton et al. 1998,
Bowles and Gintis 2002a, 2002b, Fehr and Gachter 2000b, Fehr et al. 2002,
MacIntyre 2004, Rabin 1993). Though resentment of unfairness is arguably
universal, the definition of what is fair differs culturally (Henrich 2000, Henrich et
al. 2001), as supported by the above discussion of reciprocal norms in Papua New
Guinea.
Altruistic punishment in the third-party justice game also seems at first to
conflict with both standard evolutionary and economic theory as well. But while
punishment is contrary to (economic) predictions of individuals as absolute utility
maximizersutterly self-interested individuals who will maximize utility regardless
of the utility of other individuals in the populationit may be explained in the
context of relative utility maximization (Tracer 2003). Individuals evaluate their own
fitness in reference to the fitness of others and choose the best strategy under the
given conditions in order to maximize their own fitness. So by sacrificing 20% of his
utility in the third-party game, a punisher decreases the utility of Player A by 30% or
more39. For example, if Player A proposes an offer of K0, the proposed payoffs are
Player A: 10 Player B: 0 Player C: 5.
Should Player C decide to punish, the payoffs become
39 K.3 punishment of a K.10 pay-out is 30%; K.3 punishment of a K9 pay-out is 33.3%; K3 punishment
of a K8 pay-out is 37.5%; etc..
88


Player A: 7 Player B: 0 Player C: 4.
Relative to the original Player A to Player C ratio of 10:5, the after-punishment ratio
of 7:4 is much more favorable to Player C. Thus what seems like altruistic
punishment is actually self-interested punishment as a means to produce a more even
distribution of utility (at least where Players A and C are concerned).
More difficult to explain, however, is the compensation of Player B. Luck
alone determines the payoff for inert Player B; he must rely on the benevolence of
either Player A or Player C for any pay-out, much less a large one. But Player C
should never step in to boost the payoff of any player as it contradicts both absolute
and relative fitness. Doing so may be evidence of pure altruism; though theories of
unconditional altruism do not entail punishment, they may explain why players
compensate. More likely however, are theoretical explanations about fairness and
justice.
Fairness theories may further help to explain behavior in the third-party
justice game. Though Rabins (1993) theory of faimess-of-intent may help explain
rejections (altruistic punishment) in ultimatum games, it fails to explain third-party
behavior here because proposer-action does not affect the payoff of the third-party.
Models of equity and inequality aversion (Bolton and Ockenfels 2000, Fehr and
Schmidt 1999 respectively) do give insight into the motivations observed in the
third-party justice game. Though it is readily plausible that inequality-averse third-
parties might take away money (i.e., punish) from co-players to make payoffs more
equitable, it seems less feasible that they might add to the payoffs of co-players (i.e.,
compensate) at a cost to self just to distribute payoffs more evenly. However, further
examination of qualitative results adds to the explanatory strength of these models.
Though no Player C took action for offers at or above equity (K5-K10),
quantitative data are misleading. When asked to consider a hyper-fair (>50%) offer,
Players C frequently expressed dissatisfaction for the inability to take away money
89


from Player B or add money to the payoff of Player A40. This suggests that more than
a drive to punish or compensate, Player C wanted to ensure that payoffs were evenly,
or close to evenly, distributed between all three players, even though this required
taking a loss of 20%-40%. Thus Players C seem to be concerned with the status of
their payoff relative to the other players, not just in maximizing self-utility (by
reducing the endowment of players with larger endowments) but also in maximizing
the utility of poorer co-players (by compensating Player B, and asking if it is
possible to compensate hyper-fair Player A). Fehr and Schmidt (1999) assert that
individuals who are inequality-averse will make sacrifices to both decrease the
endowment of the better-off and increase the endowment of the worse-off. This
approach requires a refinement of the traditional economic view of absolute utility
towards one that acknowledges an individuals recognition of self-utility with
reference to that of peers (i.e., individuals are both self- and other-regarding), not
unlike the evolutionary idea of relative fitness emphasized by Tracer (2003).
Moreover, it may also suggest that a sense of distributive justice drives game
behavior, with the goal of fair distribution of material gains. If indeed Au players
utilize both compensation and punishment as a means to facilitate even distribution
of utility, because of inequality aversion, cultural proclivities, or sense of justice, it
highlights the importance of both altruistic punishment and altruistic rewards for the
maintenance of cooperation (Fehr and Rockenbach 2003).
In addition to inequality aversion, culture may also influence third-party
behavior. The proliferation and strength of reciprocal relationships not unlike
40 For example, one third-party who obviously understood the rules of the game very well prior to
entering the game enclosure (unlike many), came in and immediately stated that he knew exactly what
he was going to do: take the K.5 allotted to him and walk away. However, when he was told that the
randomly matched Player A offer to which he might respond was an astoundingly hyper-generous K.9,
he was at first shocked at the generosity, and then determined to reward Player A for his hyper-
generosity by adding to his small Kl payoff. When told that he could not add to the payoff of the
proposer (Player A), he wanted to take money away from the K9 payoff of Player B. When told that
he could not do that either, he left the game-room as if defeated, upset that he could not more evenly
distribute the payoffs from his game.
90


Trivers reciprocal altruism may influence third-party action so that village-mates
playing the game receive a fair, or fairer than proposed, share. This may be
compounded by both the high degree of relatedness of villagers, as well as the lack
of anonymous interactions in everyday life. Individuals might imagine that their
anonymous co-players could very well be a child, sister, parent, friend, or clan
member who might be able to return the favor in the future. Or, in a culture where
anonymous interactions do not exist, players may be unable to trust experimenter
claims that the game is a truly anonymous exchange, and assume that, should player
identities be found out, game behavior could impact future interactions. This could
also generate generosity of offers as well as promote compensation in order to both
self-protect and ensure social bonds.
Finally, customary practices for the treatment of crime may produce the
motivation to compensate in the game. Stemming from reciprocal norms, the
traditional justice system in Papua New Guinea is both restorative and retributive,
while the formal system is decidedly retributive. In other words, paying
compensation to victims and their families or clans makes up part or all of the
punishment of transgressors. In fact, the enactment of the Compensation Act of 1991
was a formal attempt to marry the restorative, customary justice system and the
Western, retributive system that has been imposed since the early 20th century under
colonial rule. It is likely that a sense of justice based on the customary system of
restoration influences game behavior to promote compensation by third-parties.
However, these third-parties punished and compensated at precisely the same rates,
while a few players chose to both punish and compensate. These results suggest that
both the retributive and restorative systems are equally important. Indeed, the
responses to hypothetical vignettes further corroborate the idea that a combination of
retributive and restorative justice is favored. However, the higher female tendency to
compensate or both compensate and punish is very interesting. Though inconclusive
because of such a small sample size, this may suggest that women are less selfish
91


than men; that they are more inequality averse; and/or that they may favor a
combination of restorative and retributive justice over pure restorative justice, but
restorative justice over pure retributive justice. This stands in opposition to the
tendency of men to favor retributive justice over restorative justice, but restorative
justice over a combination of the two. Perhaps this difference is in fact explained by
education, as seen through gender. As discussed in Chapter 5, men are better
educated than women. Exposure to Western ideas and the retributive (based on
imprisonment) justice system that formally presides over PNG may prompt men to
favor punishment over restoration. Again, a large, gender-balanced sample that
controls for education might elucidate whether this possibility is a reality.
92


7. Conclusion and Future Directions
7.1 Conclusion
Ubiquitous social norms shape and constrain human behavior. Reciprocity,
both positive and negative, has been argued to be a key enforcement mechanism of
the social norm of altruistic cooperation (Fehr and Gachter 1998). A refinement of
this statement is that while negative reciprocity (altruistic punishment) enforces the
social norm of altruism, positive reciprocity reinforces this social norm.
Neither standard evolutionary theory nor game theory built on neoclassical
economic theory can fully explain social preferences for altruistic cooperation and
punishment, especially in one-shot encounters. A growing body of empirical
evidence at times supporting and at times conflicting with existing theory produces a
confusing mass of overlapping descriptions and explanations. The results of this
third-party justice experiment are likewise difficult to reconcile using pure
evolutionary or economic theory, as results both support and contradict the
predictions these theories offer.
While offers of 0% in the third-party justice game among the Au seem to
support the selfishness axiom of such theory, the bimodal offers at 30% and 40%, as
well as hyper-fair offers, utterly contradict it. However, both purely selfish and
hyper-fair offers could be explained by methodological issues of game frame (e.g.,
player collusion, the pictorial example of a 0% offer, and lack of experimenter
anonymity). In addition, cultural ideas about reciprocity and the lack of anonymous
real-life interactions may also play a role in producing hyper-fair offers. Threat of
93


punishment or local ideas about fairness (i.e., generosity) mostly likely explain the
modal offers.
The high rate of altruistic punishment and compensation at 34.7% also
contradicts standard economic and evolutionary theory. Though punishment may be
explained in terms of relative fitness maximization, fairness theory, or inequality
aversion, compensation is more difficult to reconcile. Theories of inequality aversion
and distributive equality may help to explain compensation in the game, but again,
culture also probably plays a role in third-party compensation as reciprocal norms
emphasize generosity and the traditional justice system places a high premium on
restoration of wrongs to victims through compensation.
Most variables did not predispose tendencies in proposal or third-party action,
with the exceptions of 1) the inverse relationship of cash-crop income and offers, and
2) provisional (but statistically insignificant) evidence that gender impacts the
tendency to punish, compensate, or do both. Females tend to compensate or do both
more than do males. Overall however, both punishment and compensation were
chosen with equal frequency, potentially highlighting social influences from the
justice system made manifest in game behavior. Indeed qualitative results suggest
that participants believe that a mix of retributive and restorative justice best
addresses both petty and violent crime. This is likely derived from the modem PNG
justice system, a complicated combination of the customary practices of victim
compensation, customary practices of swift physical punishment, as well as the
colonization-imposed retributive system based on imprisonment.
7.1.1 On the Evolution of Altruistic Cooperation
For a complete picture of the evolution of altruistic cooperation, all of the
main theories iterated in Chapter II are necessary. Altruism likely began amongst
very close, genetically related kin, which, whether rationalized or not, augmented the
inclusive fitness of the kin-group. As cognitive processes evolved, group size
94


increased, and the genetic relatedness of individuals decreased, humans extrapolated
cooperative tendencies from kin to the new group of pseudokin (MacIntyre 2004)
with whom interactions were repeated and frequent. If cooperation was reciprocally
returned with cooperation, so be it. But if cooperation was met with defection, the
individuals who could discern the identities of the defectors or who could remember
past experiences were selected for as they could reciprocate defection with defection.
Thus tit-for-tat, or what Trivers (1971) called reciprocal altruism, was bom. But as
groups continued to grow in size as did the number of partners with whom one could
interact, more efficient mechanisms for communicating and remembering the
identity of cooperators had to evolve, making the way for indirect reciprocity and
costly signaling. An individual able to discern honest signals and ignore cheap
talk was favored by natural selection. Finally, as group size grew so large that
anonymous and one-shot interactions grew frequentthereby eliminating the
efficacy of signaling and the ability to base decisions on previous experiencea
stauncher mode of enforcement was necessary to deter defection. Altruistic
punishment, then, is not a theory that stands alone to explain the evolution of
cooperation, nor does it go beyond explication of what is necessary to maintain
cooperation among large groups. Altruistic punishment is a means toward
maintaining cooperation, reciprocal or otherwise. It is a social norm enforcer. The
results presented here show that altruistic compensation may also act as a social
norm reinforcer.
7.2 Future Directions
All players knew that the other members of their trio were fellow members of
their village. Bowles and Gintis (2002b) argue that the motivation to punish is
stronger when the identification of the group is known (even if individuals are
anonymous), so that strong reciprocity is stronger when group stability is high. By
extension, we might assume that the motivation to altruistically reward (compensate)
95


may also be stronger when group stability is high. The high degree of inter-
relatedness of village members, either genetically or reciprocally, ostensibly makes it
difficult for participants to grasp the idea of an anonymous interaction and the fact
that no future interaction will depend on behavior confined to the game. If trios were
made up of members of different villages or even different language groups with
whom interaction is minimal or non-existent, it would reduce or eliminate the
importance of such confounders, namely the inter-relatedness of participants and
anonymity, thus allowing issues of justice, fairness, and inequality aversion to come
to the forefront.
It would also be interesting to give the third-party the option to take action on
hyper-fair offers. So, in addition to punishment of the proposer and compensation of
the inert second-party, the third-party would also be able to punish inert Player B
for (the good luck of) receiving too-high a payoff and compensate Player A for his
(hyper-) generosity. This type of justice game will help elucidate whether inequality
aversion based on distributive justice (fair distribution of payoffs) drives player
behavior, or if it is indeed restorative retributive justice that impacts third-party
action.
Finally, as mentioned above, further testing with a large, gender-balanced
sample is necessary to understand if gender influences third-party action. Female
participants in this study compensated more often than men, suggesting they are
more prosocial than men. This provisionally corroborates similar findings from a
dictator (Eckel and Grossmans 1998) and trust game (Croson and Buchans 1999)
that conclude that women are more selfless than men. A possibility here might be
that because men in Papua New Guinea are better educated than women, they might
have more exposure to Western ideas including those about retributive justice. A
study controlling for education may be helpful in illuminating this possibility as well.
96


7.3 Concluding Remarks
In summary, players in the third-party justice game have heterogeneous
preferences, both in offers and in third-party actions. Unlike many ultimatum and
dictator game experiments performed in geographically and culturally diverse
locales, a non-trivial percentage of hyper-fair offers, as well as a large amount of
purely selfish (0%) offers exist. Only cash-crop income is a statistically significant
predictor of offers. Gender may predict the tendency towards compensation over
punishment, though a larger sample size is necessary to test this assertion. Though
results show that the third-party only punishes and compensates for offers less than
equality ( that players would have sanctioned and compensated for hyper-fair offers as well, if
the options were available. When comparing results to existing theory, punishment
may be explained by the maximization of relative fitness, by competition, or by
fairness. Compensation may be incompletely explained by models of inequality
aversion, but perhaps is best explained by a high degree of reciprocal relations, of
inter-relatedness of participants, and the customary restorative justice system.
Generalizations from theory predict that in social dilemmas, defection or non-
cooperation produces a fitness advantage for selfish individuals relative to those who
cooperate (Schroeder et al. 2003), altruistically punish, and altruistically compensate.
However, in cultures where reciprocal exchange and compensation are valued in
daily interactions, cooperators are actually obeying and reinforcing social norms.
Thus, while other-regarding behavior may seem to flout self-interest in the short-
term, obeying norms may actually be considered to be self-interested as it prevents
sanctions and bolsters cooperative relationships that provide fitness advantages in the
long-run. In conclusion, results suggest that culture and local ideas about fairness and
justice play a much more active role in producing economic behavior than
conventional theory would allow.
97


APPENDIX A
Map and Photographs
(http://www.odci.gov/cia/publications/factbook/geos/pp.htmn
Arrow indicates the approximate location of Anguganak.
A traditional-style house. Anguganak, PNG.
A more modem, woven-walled house atop stilts.
Bogasep, PNG.
98


An even more modem house with corrugated roof.
Anguganak, PNG.
Photograph A.4
Mama Opa scraping sago.
Anguganak, PNG.
Photograph A.5
Sago scraping with a
bamboo hammer.
Anguganak, PNG.
Photograph A.6
Meini washing sago,
using her hand-made
bamboo processor.
Anguganak, PNG.
99


A typical meal of sago jelly (left) and greens
(right). The greens here were stewed with tinned
fish. Anguganak, PNG.
Community school (left). Anguganak, PNG.
The author conducting pre-game interviews.
Game enclosure (right). Winaluk, PNG.
Photo courtesy of David Tracer.
Photograph A.8
Market day at the Station. In the background
are the hangar (left) and trade-store (right).
Angugnak, PNG.
Sarah grooming Janet. Janet pinching lice.
Angugnak, PNG.
The principal investigator, David Tracer,
conducting the game with a third-party
participant. The three pieces of paper hold
the amount of money in coins allotted to
each player.
Brugap, PNG.
100


APPENDIX B
Script, Data Sheet (Pre-game Interview), Post-game Interview and Vignettes
1.1 Introductory Comments
1. Tenkyu olgeta long kam long hia tude. Dispela wok (em i olsem liklik pilai o
gem) em bai kisim longpela taim liklik kain olsem sikispela haua samting olsem
na sapos yu no inap stap long dispela longpela taim, yu mas tokaut nau. Pastaim
bipo mi wokim dispela wok, mii laik givim yu liklik toksave long wanem kain
saating mi mekim na wane mol lo na pasin yu mas bihainim long wokim dispela
wok.
Translation:
Thank you all for coming here today. This research (work) (which is like a little
game) will take a pretty long time to playmaybe six hours or soso if you cannot
stay such a long time, please say so now. Before we begin, Id like to give you a
little talk about what well be doing today and the things youll have to know before
participating.
2. Bai yumi wokim wanpela pilai wantaim mani. Long dispela pilai yu ken kisim
sampela mani na kisim i go long haus bilong yu na usim long laik bilong yu. Yu
mas save, dispela mani em ino mani bilong Daavid, nogat. Em i manii bilong
wanpela Uni (skul) is tap long Amerika na ol i bin givim bilong wokim dispela wok.
Bihain bai dispela wok is tap long buk. I gat planti ol wokman bilong Uni ol i go aut
long planti hap graun na long planti arapela kantri na mekim wankain wok. Bihain
bai mipela mumutim olgeta wok na raitim wanpela buk.
Translation:
You and I will be playing with money. When you play, you have the chance to
earn money which you can take to your house and use as you like. You must know
that this money does not belong to David (Tracer, primary investigator). It is money
that belongs to a school (university) in the U.S. and it has been given to do this
research. Later, (I will write about) this work in a book. There are many other
workers from universities who are going out to various countries and doing the same
kind of work. Later well put all the work weve done together and write a book.
101


3. Narapela samting mi laik tokim yu dispela em i wok tru tru, e mi no trik na em i
no samting bilong paulim o giamanim yu.
Translation:
Another thing Id like to talk to you aboutthis is real research (work), Im not
tricking you or pulling a joke on you.
4. Bipo yumi statim dispela wok, mi laik tokim yu bikpela samting. Taim yuk am
long hia, yu no bin save wanem kain wok mi laik mekim. Sapos yu no gat laik long
wokim, em orait, yu ken i go. Na sapos pilai i stat pinis na yu les long wokim, stile
m i orait, yu ken i go.
Translation:
Before we start, Id like to talk to you about something important. When you came
here (today), you didnt know what kind of work Id wanted to do (with you). If you
dont like this work, its alright, you can go. And if, after we start playing, you dont
like the game, its still alright, you can go.
5. Narapela samting: Bai wanwan maneri kam insait na wokim dispela pilai na
bihain bai yu go aut na wet longpela taim liklik. Taim yu stap na wet yu ken toktok
long olgeta samting tasol long dispela pilai yu bin mekim pinis yu no ken toktok.
Sapos yu tokaut long wanem kain saamting yu bin mekim long pilai, yu ken
bagarapim dispela wok.
Translation:
Another thing: one by one you (men and women) will come inside, play the game,
and got outside and wait for awhile. While waiting, you can talk (to each other)
about anything except about the game and what you did when you played. If you
talk about this kind of thing (or), what you did in the game, you will really mess up
this research.
6. Las samting e mi olsem: bilong helpim mi long wokim dispela wok bai olgeta
wanwan manmeri kisim 2 kina. Em it ok tenkyu tasol em i no stap insait long
pilai. Taim mi givim long yu, yu ken putim long sampela hap, e mi bilong yu nau.
Translation:
One last thing: for helping me with this research (work) I will give each of you 2
Kina. Its a just a thank youit doesnt have to do with playing (the game). When I
give it to you, you can put it somewhere (pocket/bag), it belongs to you now.
102


1.2 Justice Game Script
1. Long wokim dispela pilai, mi makim 3-pela manmeri, na bai 3-pela wok
wantaim. [Olgeta 3-pela ol i bilong hia (dispela pies), tasol mi no inap kolim nem
tru bilong ol na mi no inap tokaut husat i bin wok wantain husat narapela. Nogat.
Na bihain tu, taim dispela wok i pinis, em bai stap olsem, mi no inap tokaut husat i
bin wokim pilai wantaim husat narapela na em bai stap olsem oltaim.
Translation:
To play the research (game), I will make groups of three (men and/or women), and
the group of three will play together. The three people will all be from here (this
place/village), but I wont use their real names and I wont tell anyone (playing) who
the other players (members of the group) are (who played the game together). No.
And also later, when the work is finished, I wont tell who played the game with any
other personand Ill keep it a secret always.
2. Orait, dispela 3-pela mi makim long wokim pilai wantaim, bai mi kolim wanpela
manmeri namba wan, narapela manmeri namba tu, na narapela manmeri namba
tri. Na bai mi givim tenpela K1 i go long manmeri namba wan na namba tu.
Namba wan mas tingim pastaim na tokim mi olsem wanem e mi laik brukim o tilim
dispela tenpela kina namel long en na namba tu. Namba wan em i ken salim KO na
em yet holim olgeta K10 o em salim K1 i go long namba tu na em yet holim K9, o
em salim K2 na em yet holim K8, K3...i go inap e mi salim K10 olgeta i go long
namba tu na em yet bai kisim nogat olgeta. Orait, nau namba wan na namba tu ol no
inap kisim kina yet. Ol i mas wet liklik inap namba tri em i pilai, na bihain bai mi
tokim ol hamas kina ol i kisim.
Translation:
Alright, this trio that I make for the game, I will call one person person number
one (PI), another person number two (P2), and the other person person number
three (P3). And I will give ten K1 (coins) to PI and P2. PI must first thing and then
say how he/she would like to divide up the K10 between himself and P2. PI can
send KO and hold all K10, or he can send K1 to P2 and hold K9, or send K2 and
hold K8, K3 and hold K7, K4......all the way until he can send all K10 to P2 and
keep nothing at all. Alright, now PI and P2 are not finished yet (cant take the
money yet). They must wait a little for P3 to play, and later I will tell them (PI and
P2) how much Kina they will take.
3. Namba tri em bai pilai olsem: Bai mi givim faipela K1 long namba tri. Pastaim
mi tokim em hamas namba wan i bin salim i go long long namba tu na hamas em yet
holim na namba tri e mi ken wokim kainkain samting: (1) em i ken peim K1 bilong
103


rausim (tekwe) 3-pela kina bilong namba wan, (2) e mi ken peim K1 bilong skruim
(addim) 3-pela kina i go long namba tu, (3) e mi ken peim K2 bilong rausim/tekwe
3-pela kina bilong namba wan na givim i go long namba tu, o (4) namba tri em i ken
holim olgeta faipela kina bilong em yet na larim olgeta samting (kina bilong namba
wan na namba tu) i stap wankain tasol.
Translation:
P2 will play like this: I will give five K1 (coins) (a total of K5) to P3. First I will tell
him/her how much PI sent to P2 and how much PI kept; and P3 can do these things:
(1) he can pay K.1 to take away K3 (that belong to) from PI, (2) he can pay K1 to
add K3 to P2, (3) he can pay K2 to take away K3 from PI and give them to P2, or
(4) P3 can keep the five Kina that belong to him and leave it (the kina that belong to
PI and P2 as decided by PI) just as it is.
104


1.3 Standardized Examples
(Used in the general assembly with pictorial representations.)
1.3.1 PI: Keep K3, Send K7 P3: a) nothing b) K1 to take away c) K1 to add
d) both Answers = a) PI: 3, P2:7, P3: 5
b) PI: 0, P2: 7, P3: 4
c) PI: 3, P2: 10, P3: 4
d) PI: 0, P2:10, P3: 3
1.3.2 PI: Keep K7, Send K3 P3: a) nothing b) K1 to take away c) K1 to add
d) both Answers= a) PI: 7, P2: 3, P3: 5
b) PI: 4, P2: 3, P3: 4
c) PI: 7, P2: 6, P3: 4
d) PI: 4, P2: 6, P3: 3
1.3.3 PI: Keep K5, Send K5 P3: a) nothing b) K1 to take away c) K1 to add
d) both Answers= a) PI: 5, P2: 5, P3: 5
b) PI: 2, P2: 5, P3: 4
c) PI: 5, P2: 8, P3:4
d) PI: 2, P2: 8, P3: 3
105


P3: a) nothing b) K1 to take away c) K1 to add
1.3.4 PI: Keep K10, Send K0
d) both
Answers= a) PI: 10, P2: 0, P3: 5
b) PI: 7, P2: 0, P3: 4
c) PI: 10, P2: 3, P3: 4
d) PI: 7, P2: 3, P3: 3
106


1.4 Testing
(Used after Player 1 entered the game enclosure to check understanding.)
1.4.1 PI: Keep K6, Send K4 P3: a) nothing b) K1 to tekwe c) K1 to add d)
both Answers= a) PI: 6, P2: 4, P3: 5
b) PI: 3, P2: 4, P3: 4
c) PI: 6, P2: 7, P3: 4
d) PI: 3, P2: 7, P3: 3
1.4.2 PI: Keep K4, Send K6 P3: a) nothing b) K1 to tekwe c) K1 to add d)
both Answers= a) PI: 4, P2: 6, P3: 5
b) PI: 1, P2: 6, P3: 4
c) PI: 4, P2: 9, P3: 4
d) PI: 1, P2: 9, P3: 3
Extra if needed:
1.4.3 PI: Keep K8, Send K2 P3: a) nothing b) K1 to tekwe c) K1 to add d)
both Answers= a) PI: 8, P2: 2, P3: 5
b) PI: 5, P2: 2, P3: 4
c) PI: 8, P2: 5, P3: 4
d) PI: 5, P2: 5, P3: 3
107


1.5 Player Data Sheet
ID Code________________Village _______________________________Sex______________
Kolim nem bilong vu? (name)____________________________________________________
Kolim papa nem bilong yu?______(fathers name, which acts as a surname)________
Yu save wanem yia mama karim yu o hamas krismas bilong yu? (age)
Yu marit o nogat? (married or not) (if yes) Wanpela meri tasol o moa? (# of wives)
Hamas pikinnini bilong yu? ___________________(# of children)__________________
Yu skul o nogat? (attended school) (If yes) Yu pinisim wanem gred?(highest grade)
Yu save go long lotu o nogat? (attend church)__________
(if Y) Hamas taim long wanwan mun? (# times/mo.)
Hamas gaden kaikai yu planim na gat nau?______(# of gardens)___________________
Yu save planim kopi? (vou plant coffee) Kakao? (cocoa) Vanila?________(vanilla)
Yu save salim na kisim hamas kina long wanwan mun long k, k, o vanilla?________
_________________(from selling these cash crops, how much do vou make a month?)
Yu save wok bilong kisim kina sampela taim o stap long ples/wokim gaden tasol?_
___________________(Do vou work for wages some of the time or work vou garden?)
(If work) Bilong wok, yu save kisim hamas kina long wanwan mun? (how much do vou earn
per month?)____________________________________________________________________
*____________________________________________________________________________________
*
Role: PI P2 P3
Associated Player IDs: PI_____________P2______________P3_______________
PI Offer_______________________________
P3 Action______________________________
Payouts: PI_____________P2______________P3_______________
Comments (level of understanding, etc):______________________________________________
108


1.6 Post-Game Questions and Vignettes
For Players 1: Long tingting bilong yu, hamas kina ol arapela manmeri i bin
givim/salim i go long namba tu? (What do you think, how much [kinal did
other people give/send to Player 2?)_______________________________________
Bilong wanem?_______(Why?)
For Players 3: Long tingting bilong yu, ol arapela namba tri bin peim kina
bilong rausim sampela kina bilong namba wan o skruim sampela kina long
namba tu? (What do you think, did the other Players 3 pay kina to take away
king from PI or to add kina to P2?)_____________________________________________
Bilong wanem?_______(Why?)
Vignettes
#1. Theft of Found Kina A man finds K10 on the ground when walking through the
bush. He takes it home, hides it, locks his house, and leaves. When he returns,
someone has broken in and stolen it.
a) the thief is jailed (follow up: what length jail term is fairest?)
b) the thief is fined, and the fine goes to the state (follow up: what size fine is
fairest?)
c) the victim receives financial compensation from the state (follow up: what size
compensation is fairest?)
d) the victim receives financial compensation from the thief (follow up: what size
compensation is fairest?)
e) the thief is killed (follow up: by whom?)
f) other (suggest_______________)
109


#2. Theft of Earned Kina A man hides K10 that he earned in his house, locks the
house, and leaves. When he returns, someone has broken in and stolen the K10.
a) the thief is jailed (follow up: what length jail term is fairest?)
b) the thief is fined, and the fine goes to the state (follow up: what size fine is
fairest?)
c) the victim receives financial compensation from the state (follow up: what size
compensation is fairest?)
d) the victim receives financial compensation from the thief (follow up: what size
compensation is fairest?)
e) the thief is killed (follow up: by whom?)
f) other (suggest_______________)
#3. Accidental Killing/Manslaughter While driving, a man hits another man with
his car and kills him.
a) the driver is jailed (follow up: what length jail term is fairest?)
b) the driver is fined, and the fine goes to the state (follow up: what size fine is
fairest?)
c) the victim's clan receives financial compensation from the state (follow up: what
size compensation is fairest?)
d) the victim's clan receives financial compensation from the driver (follow up: what
size compensation is fairest?)
e) the driver is killed (follow up: by whom?)
f) other (suggest______________)
110


4. Homicide Two friends get into a fight. One becomes so angry he takes a gun and
fatally shoots his friend.
a) the shooter is jailed (follow up: what length jail term is fairest?)
b) the shooter is fined, and the fine goes to the state (follow up: what size fine is
fairest?)
c) the victim's clan receives financial compensation from the state (follow up: what
size compensation is fairest?)
d) the victim's clan receives financial compensation from the shooter (follow up:
what size compensation is fairest?)
e) the shooter is killed (follow up: by whom?)
f) other (suggest______________)


APPENDIX C
Human Subjects Review Board Approval
University of Colorado at Denver and Health Sciences Center
Human Subjects Research Committee Institutional Review Board
Downtown Denver
Campus Box 120, P.O. Box 173364
Denver, Colorado 80217-3364
Phone: 303-556-4060, Fax: 303-556-5855
DATE:
TO:
FROM:
SUBJECT:
February 18, 2005
David Tracer
Deborah Kellogg, HSRC Chair
/meirgr
Human Subjects Research Protocol #2005-083 Prosociality and Justice: A
Cross-Cultural Experimental Study
Your protocol, with changes, has been approved as non-exempt and should pose no more than
minimal risk. This approval is good for up to one year from this date.
Your responsibilities as a researcher include:
If you make changes to your research protocol or design you should contact the
HSRC.
You are responsible for maintaining all documentation of consent. Unless
specified differently in your protocol, all data and consents should be
maintained for three years.
If you should encounter adverse human subjects issues, please contact us
immediately.
If your research continues beyond one year from the above date, contact the
HSRC for an extension.
The HSRC may audit your documents at any time.
Good Luck with your research.
112


BIBLIOGRAPHY
Alexander, J. McKenzie
2000 Evolutionary Explanations of Distributive Justice. Philosophy of Science
67:490-516.
Alexander R.D.
1979 Darwinism and Human Affairs. Seattle: University of Washington Press.
Austen-Smith, David and J.S. Banks
2002 Costly Signaling and Cheap Talk in Models of Political Influence. European
Journal of Political Economy 18:263-280.
Axelrod, R.
1984 The Evolution of Cooperation. New York: Basic Books, Inc.
Axelrod Robert and William D. Hamilton
1981 The Evolution of Cooperation. Science 211:1390-1396.
Balasko, Yves
1988 Foundations of the Theory of General Equilibrium: Economic Theory,
Econometrics, and Mathematical Economics. Orlando: Academic Press Inc.
Banks, Cyndi
1998 Custom in the Courts: Criminal Law (Compensation) Act of Papua New
Guinea. British Journal of Criminology 38(2):299-317.
Barnett, Randy E.
1977 Restitution: A New Paradigm of Criminal Justice. Ethics 87:279-301.
Bartholdi, John J. Ill, C.A. Butler, and M.A. Trick
1986 More on the Evolution of Cooperation. The Journal of Conflict Resolution
30( 1): 129-140.
Bennet, Christopher
2002 The Varieties of Retributive Experience. The Philosophical Quarterly
52(207): 145-163.
113


Berg, Joyce, J. Dickhaut, and K. McCabe
1995 Trust, Reciprocity, and Social History. Games and Economic Behavior
10:122-142.
Bolton, Gary E., J. Brandts, E. Katok, A. Ockenfels, and R. Zwick
N.d. Testing Theories of Other-Regarding Behavior. Discussion papers on
Strategic Interaction 2002-43, Max Planck Institute of Economics, Strategic
Interaction Group.
Bolton, Gary E., E. Katok, and R. Zwick
1998 Dictator Game Giving: Rules of Fairness Versus Acts of Kindness.
International Journal of Game Theory 27:269-299.
Bolton, Gary E., and A. Ockenfels
2000 ERC: A Theory of Equity, Reciprocity, and Competition. American
Economic Review 90(1): 166-193.
Bowles, Samuel, and H. Gintis
2002a Homo reciprocans. Nature 415:125-128.
2002b Social Capital and Community Governance. The Economic Journal
112:F419-F436.
2004 The Evolution of Strong Reciprocity: Cooperation in Heterogeneous
Populations. Theoretical Population Biology 65:17-28.
Boyd, Robert, H. Gintis, S. Bowles, and P.J. Richerson
2002 The Evolution of Altruistic Punishment. Proceedings of the National
Academy of the Sciences 100(6):3531-3535.
Boyd, Robert and P.J. Richerson
2005 The Origin and Evolution of Cultures. Oxford: University Press.
Bradley, B.J.
1999 Levels of Selection, Altruism, and Primate Behavior. Quarterly Review of
Biology 74(2): 171-194.
Braithwaite, John
1996 Restorative Justice and a Better Future. In A Restorative Justice Reader. G.
Johnstone, ed. Pp.83-107. Portland: Willan Publishing.
114


Brosnan, S., and F.B.M. de Waal
2002 A Proximate Perspective on Reciprocal Altruism. Human Nature 13(1): 129-
152.
Burling, Robbins
1962 Maximization Theories and the Study of Economic Anthropology. American
Anthropologist 64(4): 802-821.
Camerer, Colin
1999 Behavioral Economics: Reunifying Psychology and Economics. Proceedings
of the National Academy of the Sciences 96:10575-10577.
2003 Behavioral Game Theory: Experiments in Strategic Interaction. New York:
Russell Sage Foundation.
Cameron, L.
1995 Raising the Stakes in the Ultimatum Game: Experimental Evidence from
Indonesia. Discussion Paper, Princeton University.
Central Intelligence Agency (CIA)
2005 CIA World Factbook. Electronic document,
www.odci.gov/cia/publications/factbook/geos/pp.html, accessed July 15.
Chapais, B.
2001 Primate Nepotism: What is the Explanatory Value of Kin Selection?
International Journal of Primatology 22(2):203-229.
Chapais, B., L. Savard, and C. Gauthier
2001 Kin Selection and the Distribution of Altruism in Relation to Degree of
Kinship in Japanese Macaques (Macaca fuscata). Behavioral Ecology and
Sociobiology 49:493-502.
Chamess, Gary, and M. Rabin
2002 Understanding Social Preferences with Simple Tests. The Quarterly Journal
of Economics Aug:817-869.
Clark, Kenneth, and M. Sefton
2001 The Sequential Prisoners Dilemma: Evidence on Reciprocation. The
Economic Journal 111:51-68.
115


Cohen, Ronald L.
2001 Provocations of Restorative Justice. Social Justice Research 14(2):209-232.
Conroy, G.
2005 The Emergence of Culture and the Origins of the Genus Homo. In
Reconstructing Human Origins: A Modem Synthesis. 2nd edition. Pp 294-343.
New York: WW Norton.
Cook, Scott
1966 The Obselete Anti-Market Mentality: A Critique of the Substantive
Approach to Economic Anthropology 68(2):323-345.
Croson, Rachel, and N. Buchan
1999 Gender and Culture: International Experimental Evidence from Trust Games.
Gender and Economic Transactions 89(2):386-391.
Dalton, George
1969 Theoretical Issues in Economic Anthropology. Current Anthropology 10(1):
63-102.
Daly, Kathleen
2000 Revisiting the Relationship Between Retributive and Restorative Justice. In
Restorative Justice: From Philosophy to Practice. H. Strang and J. Braithwaite,
eds. Dartmouth: Aldershoot.
Daly, M., and M. Wilson
1978 Sex, Evolution, and Behavior. 2nd edition. Belmont: Wadsworth Publishing
Company.
Darlington, P.J. Jr.
1972 Nonmathematical Models for Evolution of Altruism, and for Group
Selection. Proceedings of the National Academy of the Sciences 69(2):293-
297.
1978 Altruism: Its Characteristics and Evolution. Proceedings of the National
Academy of the Sciences 75(l):385-389.
Darwin, Charles
1859 The Origin of Species by Natural Selection. London: John Murray.
116


Dawkins, R.
1976 The Selfish Gene. New York: Oxford University Press.
De Waal, F.
1991. The Chimpanzees Sense of Social Regularity and its Relation to the Human
Sense of Justice. American Behavioral Scientist 34:335-349.
Debreu, Gerard, and H. Scarf
1963 A Limit on the Core of an Economy. International Economic Review 4:235-
246.
Diekmann, Andreas
2004 The Power of Reciprocity: Fairness, Reciprocity, and Stakes in Variants of
the Dictator Game. Journal of Conflict Resolution 48(4):487-505.
Eckel, Catherine C., and P.J. Grossman
1998 Are Women Less Selfish than Men?: Evidence from Dictator Experiments.
The Economic Journal 108(May):726-735.
Falk, Armin, E. Fehr, and U. Fischbacher
2003 On the Nature of Fair Behavior. Economic Inquiry 41(l):20-26.
Fehr, Ernst and U. Fischbacher
2002a The Nature of Human Altruism. Nature 425:785-791.
2002b Why Social Preferences Matterthe Impact of Non-selfish Motives on
Competition, Cooperation, and Incentives. The Economic Journal 112:C1-C33.
2004a Social Norms and Human Cooperation. TRENDS in Cognitive Sciences
8(4):185-190.
2004b Third-party Punishment and Social Norms. Evolution and Human Behavior
25:63-87.
Fehr, Ernst, U. Fischbacher, and S. Gachter
2002 Strong Reciprocity, Human Cooperation and the Enforcement of Social
Norms. European Economic Review 42:845-859.
117


Fehr, Ernst, U. Fischbacher, and E. Tougareva
2002 Do High Stakes and Competition Undermine Fairness? Evidence from
Russia. Working Paper No. 120. Institute for Empirical Research in Economics,
University of Zurich.
Fehr, Ernst, and S. Gachter
1998 Reciprocity and Economics: the Economic Implications of Homo
Reciprocans. European Economic Review 42:845-859.
2000a Cooperation and Punishment in Public Goods Experiments. The America
Economic Review. 90(4):980-994.
2000b Fairness and Retaliation: the Economics of Reciprocity. Journal of
Economic Perspectives 14:159-181.
2002 Altruistic Punishment in Humans. Nature 415:137-140.
Fehr, Ernst, and J. Henrich
In press Is Strong Reciprocity a Maladaptation? In The Genetic and Cultural
Evolution of Cooperation. P. Hammerstein, ed. Cambridge: MIT Press.
Fehr, Ernst, and B. Rockenbach
2003 Detrimental Effects of Sanctions on Human Altruism. Nature 422(13): 137-
MO.
2004 Human Altruism: Economic, Neural, and Evolutionary Perspectives. Current
Opinion in Neurobiology 14:784-790.
Fehr, E., and K. Schmidt
1999 A Theory of Fairness, Competition, and Cooperation. The Quarterly Journal
of Economics 817-868.
Fisher, Arthur
1992 Sociobiology: Science or Ideology? Society 29(5):67-80.
Fishman, Michael A.
2003 Indirect Reciprocity among Imperfect Individuals. Journal of Theoretical
Biology 225:285-292.
Fleagle, J.G.
1999 Primate Adaptation and Evolution. 2nd edition. San Diego: Academic Press.
118


Foucault, Michel
1977 Discipline and Punish: The Birth of the Prison. Alan Sheridan, trans. New
York: Pantheon Books.
Fowler, James H., T. Johnson, and O. Smirnov
2004 Egalitarian Motive and Altruistic Punishment. Nature 433:E1.
Gintis, H., and S. Bowles
2003 The Origins of Human Cooperation. In The Genetic and Cultural Origins of
Culture. P. Hammerstein, ed. Pp. 1-17. Cambridge: MIT Press.
Gintis, H., E.A. Smith, and S. Bowles
2001 Costly Signaling and Cooperation. Journal of Theoretical Biology 213:103-
119.
Gintis, Herbert
2000 Strong Reciprocity and Human Sociality. Journal of Theoretical Biology
206:169-179.
Gintis, Herbert, S. Bowles, R. Boyd, and E. Fehr
2003 Explaining Altruistic Behavior in Humans. Evolution and Human Behavior
24:153-172.
Gouldner, Alvin W.
1960 The Norm of Reciprocity: a Preliminary Statement. American Sociological
Review 25:161-178.
Gowdy, John, and I. Seidl
2004 Economic Man and Selfish Genes: the Implications of Group Selection for
Economic Valuation and Policy. Journal of Socio-Economics 33:343-358.
Griset, Pamela L.
1991 Determinate Sentencing: the Promise and the Reality of Retributive Justice.
Albany: State University of New York Press.
Hamilton, W.D.
1964. The Genetical Theory of Social Behavior: I and II. Journal of Theoretical
Biology 7:1-52.
1972 Altruism and Related Phenomena, Mainly in Social Insects. Annual Review
of Ecology and Systematics 3:193-232.
119


1975 Innate Social Aptitudes of Man: an Approach from Evolutionary Genetics. In
Biosocial Anthropology. R. Fox, ed. Pp. 133-153. London: Malaby Press.
Hampton, Jean
1984 The Moral Education Theory of Punishment. Philosophy and Public Affairs
13:208-238.
Henrich, J., R. Boyd, S. Bowles, C. Camerer, E. Fehr, H. Gintis, and R. McElreath
2001 In Search of Homo Economicus: Behavioral Experiments in 15 Small-scale
Societies. The American Economic Review 91(2):73-78.
Henrich, Joseph.
2000 Does Culture Matter in Economic Behavior? Ultimatum Game Bargaining
among the Machiguenga of the Peruvian Amazon. The American Economic
Review 90(4):973-979.
2001 Cultural Group Selection, Coevolutionary Processes and Large-scale
Cooperation. Journal of Economic Behavior & Organization 53:3-35.
Henrich, Joseph, and R. Boyd
2001 Why People Punish Defectors: Weak Conformist Transmission Can Stabilize
Costly Enforcement of Norms in Cooperative Dilemmas. Journal of Theoretical
Biology 208:79-89.
Henrich, Joseph, R. Boyd, S. Bowles, C. Camerer, E. Fehr, and H. Gintis, eds.
2004 Foundations of Human Sociality. Economic Experiments and Ethnographic
Evidence from Fifteen Small-Scale Societies. Oxford: University Press.
Hoffman, Elizabeth, K. McCabe, K. Shachat, and V. Smith
1994 Preferences, Property Rights, and Anonymity in Bargaining Games. Games
and Economic Behavior 7:346-380.
Hoffman, Elizabeth, K. McCabe, and V. Smith
1996 Social Distance and Other-regarding Behavior in Dictator Games. The
American Economic Review 86(3):653-660.
Holt, Charles A., and A.E. Roth
2004 The Nash Equilibrium: a Perspective. Proceedings of the National Academy
of the Sciences 101(12):3999-4002.
120


Humphrey, Nicholas
1997 Varieties of Altruismand the Common Ground Between. Social Research
64(2): 199-210.
Johnstone, Gerry, ed.
2003 A Restorative Justice Reader: Texts, Sources, Context. Portland: Willan
Publishing.
Kitcher, Philip
1993 The Evolution of Human Altruism. The Journal of Philosophy 90(10):497-
516.
Kreps, David, and A. Rubinstein
1997 An Appreciation. In Classics in Game Theory. H.W. Kuhn, ed. Pp. xi-xv.
Princeton: Princeton University Press.
LeClair, Edward E. Jr.
1962 Economic Theory and Economic Anthropology. American Anthropologist
64(6): 1179-1203.
Lewis, Gilbert
1980 Day of Shining Red. An Essay on Understanding Ritual. Cambridge:
Cambridge University Press.
MacIntyre, Ferren
2002 Was Religion a Kinship Surrogate? Journal of the America Academy of
Religion 72(3):653-694.
Marshall, Tony F.
1998 Restorative Justice: an Overview. In A Restorative Justice Reader. G.
Johnstone, ed. Pp. 46-56. Portland: Willan Publishing.
Matessi, Carlo, and S. Karlin
1984 On the Evolution of Altruism by Kin Selection. Proceedings of the National
Academy of the Sciences 81:1754-1758.
Maynard-Smith, J.
1982 Evolution and the Theory of Games. Cambridge: University Press.
121


McAndrew, Francis T.
2002 New Evolutionary Perspectives on Altruism: Multilevel-Selection and Costly
Signaling Theories. Current Directions in Psychological Science 11(2):79-82.
McGrew, W.C.
1998. Culture in Nonhuman Primates? Annual Review of Anthropology 27:301-
328.
Mohtashemi, Mojdeh, and L. Mui
2003 Evolution of Indirect Reciprocity by Social Information: the Role of Trust
and Reputation in Evolution of Altruism. Journal of Theoretical Biology
223:523-531.
Nash, John F. Jr.
1950a The Bargaining Problem. Econometrica 18:155-162.
1950b Equilibrium Points in N-person Games. Proceedings of the National
Academy of the Sciences 36:48-49.
New York Times
2005 From Kuwait: A Message of Hope. New York Times, September 14: A7.
Nowak, Martin A., K.M. Page, and K. Sigmund
2000 Fairness Versus Reason in the Ultimatum Game. Science 289(5485): 1773-
1776.
Nowak, Martin A., and K. Sigmund
1998 The Dynamics of Indirect Reciprocity. Journal of Theoretical Biology
194:561-574.
Papua New Guinea National Statistics Office (PNGNSO)
2005 Census Data. Electronic document,
http://www.nso.gov.pg/Pop_Soc_%20Stats/popsoc.htm, accessed July 15.
Polanyi, Karl, C.M. Arensberg, and H.W. Pearson, eds.
1957 Trade and Market in the Early Empires. Economies in History and Theory.
Glencoe, Illinois: The Free Press.
Quervain, Dominique J.-F., U. Fischbacher, V. Treyer, M. Schellhammer, U.
Schnyder, A. Buck, and E. Fehr
2004 The Neural Basis of Altruistic Punishment. Science 305:1254-1258.
122


Rabin, Matthew
1993 Incorporating Fairness into Game Theory and Economics. The American
Economic Review 83(5): 1281-1302.
Rabinow, Paul, ed.
1984 The Foucault Reader. New York: Pantheon Books.
Rappaport, Roy A.
1968 Pigs for the Ancestors. Ritual in the Ecology of a New Guinea People. New
Haven: Yale University Press.
Rawls, J.
1999 A Theory of Justice. Rev. ed. Cambridge: Harvard University Press.
Reed, K.E., and L.R. Bidner
2004 Primate Communities: Past, Present and Possible Future. Yearbook of
Physical Anthropology 47: 2-39.
Riechmann, Thomas
2002 Relative Payoffs and Evolutionary Spite: Evolutionary Equilibriums in
Games with Finitely Many Players. Discussion paper 260. University of
Hannover.
Rischer, Nicholas
2002 Fairness: Theory & Practice of Distributive Justice. New Brunswick:
Transaction Publishers.
Romp, Graham
1997 Game Theory. Introduction and Applications. New York: Oxford University
Press, Inc.
Roth, A.E, V. Prasnikar, M. Okuno-Fujiwara, and S. Zamir
1991 Bargaining and Market Behavior in Jerusalem, Ljubljana, Pittsburgh, and
Tokyo: an Experimental Study. The American Economic Review 81(5): 1068-
1095.
Sahlins, Marshall
1972 Stone Age Economics. New York: Aldine.
123


Schroeder, David A., J.E. Steel, A.J. Woodell, and A.F. Bembenek
2003 Justice Within Social Dilemmas. Personality and Social Psychology Review
7(4):374-387.
Seyfarth, R.M. and D.L. Cheney
1984 Grooming, Alliances, and Reciprocal Altruism in Vervet Monkeys. Nature
308:541-543.
Shapley, L.S.
1953 A Value for ^-person Games. In Classics in Game Theory. H.W. Kuhn, ed.
Pp. 69-79. Princeton: Princeton University Press.
Sigmund, Karl, E. Fehr, and M.A. Nowak
2002 The Economics of Fair Play. Scientific American, Jan.:82-87.
Silk, J.B.
2004 The Evolution of Cooperation in Primate Groups. In Moral Sentiments and
Material Interests: the Foundation of Cooperation in Economic Life. H. Gintis,
S. Bowles, R. Boyd, and E. Fehr, eds. Cambridge: MIT Press.
Sillitoe, Paul
1998 An Introduction to the Anthropology of Melanesia: Culture and Tradition.
Cambridge: University Press.
Smith, Adam
1868 [1863] An Inquiry into the Nature and Causes of the Wealth of Nations. Rev.
edition. Edinburgh: Adam and Charles Black.
Smith, M.
1976. Group Selection. Quarterly Review of Biology 51:277-283.
Strier, K.B.
2000 Evolution and Social Behavior. In Primate Behavioral Ecology. 2nd edition.
Pp. 94-134. Boston: Allyn and Bacon.
Telser, L.G.
1995 The Ultimatum Game and the Law of Demand. The Economic Journal
105(433): 1519-1523.
124


Terborgh, J. and C.H. Janson
1986 The Socioecology of Primate Groups. Annual Review of Ecology and
Systematics 17:111-135.
Tracer, David
1991 The Interaction of Nutrition and Fertility among Au Forager-Horticulturalists
of Papua New Guinea. Ph.D. dissertation, Department of Anthropology,
University of Michigan.
2003 Selfishness and Fairness in Economic and Evolutionary Perspective: an
Experimental Economic Study in Papua New Guinea. Current Anthropology
44(3):432-438.
2004 Market Integration, Reciprocity, and Fairness in Rural Papua New Guinea:
Results from a Two-village Ultimatum Game Study. In Foundations of Human
Sociality: Economic Experiments and Ethnographic Evidence from Fifteen
Small-Scale Societies. Henrich, J., R. Boyd, S. Bowles, C. Camerer, E. Fehr,
and H. Gintis, eds. Pp. 232-259. Oxford: University Press.
Trivers, R.L.
1971 The Evolution of Reciprocal Altruism. Quarterly Review of Biology 46:35-
57.
Walgrave, Lode
2004 Has Restorative Justice Appropriately Responded to Retribution Theory and
Impulses. In Critical Issues in Restorative Justice. H. Zehr and B. Toews, eds.
Pp.47-60. New York: Criminal Justice Press.
Walster, E., G.W. Walster, and E. Berscheid
1978 Equity: Theory and Research. Boston: Allyn & Bacon.
Wedekind, Claus, and M. Milinski
1996 Human Cooperation in the Simultaneous and the Alternating Prisoners
Dilemma: Pavlov Versus Generous Tit-for-Tat. Proceedings of the National
Academy of the Sciences 93:2686-2689.
Wilson, David Sloan
1983 The Group Selection Controversy: History and Current Status. Annual
Review of Ecological Systems 14:159-187.
125


Zahavi, A.
1975 Mate SelectionA Selection for a Handicap. Journal of Theoretical Biology
53(1):205-14.
1977 The Cost of Honesty (further remarks on the handicap principle). Journal of
Theoretical Biology 67(3):603-5.
Zehr, Howard
1985 Retributive Justice, Restorative Justice. In A Restorative Justice Reader. G.
Johnstone, ed. Pp. 69-82. Portland: Willan Publishing.
Zimmer-Tamakoshi, Laura
1997 The Last Big Man: Development and Mens Discontents in the Papua New
Guinea Highlands. Oceania 68:107-122.
126



PAGE 1

ALTRUISTIC PUNISHMENT AND COMPENSATION: RESULTS FROM A THIRD-PARTY JUSTICE GAME IN PAPUA NEW GUINEA by Rachel D. Foreman B.S., Sewanee: The University ofthe South, 2000 A thesis submitted to the University of Colorado at Denver in partial fulfillment of the requirements for the degree of Master of Arts Anthropology 2006 fr\Ll L ..

PAGE 2

This thesis for the Master of Arts degree by Rachel D. Foreman has been approved by David P. Tracer John Brett /""' "' / .....-------; / t Beekman 11-17> -b0\.5 Date

PAGE 3

Foreman, Rachel D. (M.A., Anthropology) Altruistic Punishment and Compensation: Results from a Third-party Justice Game in Papua New Guinea Thesis directed by Associate Professor David P. Tracer ABSTRACT The results of a third-party justice game performed among the Au of Papua New Guinea are presented. Contrary to the predictions of canonical economic and evolutionary theory, but consistent with results from previous experiments, players do not always behave selfishly: the bimodal offers are 30% and 40%, and the third-party altruistically acts (to punish, compensate, or do both) 34.7% ofthe time. Players punish and compensate equally at 13.0% of the time each, suggesting that both altruistic punishment and altruistic compensation are means of sustaining cooperation among the Au. It is argued that this is due to inequality aversion compounded by the cultural influences of 1) a structured system of reciprocal exchange that places a premium on generosity; and 2) the justice system of Papua New Guinea, a mixture of the customary restorative justice system and the Western retributive justice system, imposed on Papua New Guineans by colonial administrations. Results of this study have both theoretical and empirical implications, as they foster debate about the theories for the evolution of altruistic cooperation, and highlight the need for future testing of such theory, especially toward refining understanding of the influences of gender, fairness, and justice on game behavior. This abstract accurately represents the content oft its publication. lll

PAGE 4

DEDICATION I dedicate this thesis to Sherie, proof that altruism exists. I also dedicate this thesis to my parents, Dianne, Paul, and Katie, for your unconditional support and love, now and ever; and to Joshua, for unfaltering understanding, support, and encouragement during the cold-shower days and complaint-filled nights ofwriting.

PAGE 5

ACKNOWLEDGMENT My thanks to the John D. and Katherine T. MacArthur Foundation for funding, and to David P. Tracer for the opportunity to travel to Papua New Guinea as his research assistant on the project. Without the collaboration of Dr. Tracer and his sharing of data, this thesis would not exist. Tenkyu tru. Gratitude is also due to the Department of Anthropology, University of Colorado at Denver, for additional funding and also for support and guidance throughout my graduate studies. Special thanks to Connie Turner for all of her generosity and help the past two years. Many thanks also to Dr. John Brett and Dr. Christopher Beekman for all of their guidance and many helpful suggestions during the writing process. Finally, thanks to the people of Anguganak, Papua New Guinea for their kindness and generosity, and because of whom I stayed well fed of sweet potatoes and did not succumb to my bout of dengue fever or malaria, whichever I had.

PAGE 6

CONTENTS Figures ................................................................................................................... x Tables .................................................................................................................... xi Maps .................................................................................................................... xii Photographs ........................................................................................................ xiii Chapter 1. Introduction and Specific Aims ......................................................................... 1 1. 1 Central Question ............................................................................................. 1 I.2 Justification of the Site .................................................................................... 2 1.3 Specific Aims and Predictions ........................................................................ 4 I.4 Significance .................................................................................................... 6 I.5 Chapter Summaries ......................................................................................... 7 2. Theoretical and Empirical Background ............................................................. 9 2.I Introduction ..................................................................................................... 9 2.2 Human and Non-human Altruism ................................................................. II 2.3 Theories for the Evolution of Altruism ......................................................... 14 2.3.1 Kin Selection .............................................................................................. 14 2.3.2 Reciprocal Altruism ................................................................................... 17 2.3.3 Indirect Reciprocity and Costly Signaling .................................................. 20 VI

PAGE 7

2.3.4 Altruistic Punishment.. ............................................................................... 23 2.4 Behavioral Economics .................................................................................. 30 2.4.1 A Description and History ......................................................................... 30 2.4.2 Homo oeconomicus .................................................................................... 32 2.4.3 Games: The Prisoner's Dilemma, Dictator Game, and Ultimatum Game .. 35 2.4.4 Weaknesses of Game Theory ..................................................................... 43 2.5 Conclusion .................................................................................................... 44 3. Study Population ............................................................................................. 48 3.1 The Setting .................................................................................................... 48 3.1.1 Papua New Guinea ..................................................................................... 48 3.1.2 Anguganak, Sandaun Province .................................................................. 50 3.2 The People and Culture ................................................................................. 51 3.2.1 The Au ....................................................................................................... 51 3.2.2 Reciprocal Exchange .................................................................................. 53 3.2.3 Law and Justice .......................................................................................... 54 4. Justice ............................................................................................................. 57 4.1 Types of Justice ............................................................................................. 57 4.1.1 Retributive Justice ...................................................................................... 57 4.1.2 Restorative Justice ...................................................................................... 59 4.1.3 Distributive Justice and Fairness ................................................................ 60 4.2 Justice and Game Theory .............................................................................. 62 VII

PAGE 8

5. Methods .......................................................................................................... 64 5.1 Sample Recruitment. ..................................................................................... 64 5.2 The Game ...................................................................................................... 65 5.3 Data Collection ............................................................................................. 67 5.4 Analysis ........................................................................................................ 70 5.5 Methodological Issues ................................................................................... 71 6. Results and Discussion .................................................................................... 74 6.1 Quantitative Results ...................................................................................... 74 6.1.1 Frequencies ofVariables ............................................................................ 74 6.1.2 Player A Offers .......................................................................................... 76 6.1.3 Sanctions and Compensations .................................................................... 77 6.2 Qualitative Results ........................................................................................ 82 6.2.1 Interviews and Vignettes ............................................................................ 82 6.3 Discussion ..................................................................................................... 83 7. Conclusion and Future Directions ................................................................... 93 7.1 Conclusion .................................................................................................... 93 7.1.1 On the Evolution of Altruistic Cooperation ............................................... 94 7.2 Future Directions .......................................................................................... 95 7.3 Concluding Remarks ..................................................................................... 97 Vlll

PAGE 9

Appendix A. Map and Photographs ..................................................................................... 98 B. Script, Data Sheet (Pre-game Interview), Post-game Interview and Vignettes ................................................... 101 C. Human Subjects Review Board Approval .................................................... 112 Bibliography ...... 0000 00 oo oo oo 00 113 IX

PAGE 10

FIGURES Figure I. Education by Gender. ...................................................................................... 71 2. Education by Gender and Village ................................................................... 71 3. Distribution of Selected Means ....................................................................... 7 6 4. Vanilla and Cocoa by Village ......................................................................... 76 5. Distribution of Player A Offer (Kina) and Player C Action ............................ 78 6. Distribution of Player C Action for Offers 0%-50% ....................................... 80 7. Vignette Responses ......................................................................................... 83 X

PAGE 11

TABLES Table 1. Example of a Prisoner's Dilemma Payoff Scheme ......................................... 36 2. Role by Village and Gender Totals ................................................................ 70 3. Role by Gender and Village ........................................................................... 70 4. Frequencies ofVariables ................................................................................ 74 5. ANOVA Comparison ofVariable Means Between Villages ......................... 75 6. Frequency of Player A Offer (Kina) at Brugap, Winaluk, and Anguganak .... 77 7. Frequency of Player A Offer (Kina) and Player C Action at Brugap, Winaluk, and Anguganak ................................................... 78 8. Player C Action by Offer ............................................................................... 79 9. Player C Decisions Overall and by Gender .................................................... 80 I 0. Summary of Mean and Modal Offers in Ultimatum Games ......................... 84 XI

PAGE 12

MAPS Map Aolo Papua New Guinea 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 098 XII

PAGE 13

PHOTOGRAPHS Photograph A. I. A traditional-style house ............................................................................. 98 A.2. A more modem, woven-walled house atop stilts ......................................... 98 A.3. An even more modem house with corrugated roof ..................................... 99 A.4. Mama Opa scraping sago ............................................................................ 99 A.5. Sago scraping with a bamboo hammer ........................................................ 99 A.6. Meini washing sago, using her hand-made bamboo processor .................... 99 A. 7. A typical meal of sago jelly and greens ..................................................... 100 A.8. Market day at the Station .......................................................................... 100 A.9. Community school .................................................................................... 100 A.l 0. Sarah grooming Janet.. ............................................................................ 100 All. The author conducting pre-game interviews ........................................... 100 A.12. The principal investigator, David Tracer, conducting the game with a third-party participant ............................................................. 1 00 X Ill

PAGE 14

1. Introduction and Specific Aims 1.1 Central Question Altruism--doing an act that benefits another individual at a cost to oneselfis anomalous under the singular premise of both canonical economic and evolutionary theory. This premise predicts that individuals are self-interested maximizers of material gains. Sub-theories from varied disciplines have been advanced in the attempt to explain how altruistic cooperation could arise and be maintained among a population of egoists, and ultimately become evolutionarily stable by natural selection. These theories include inclusive fitness, reciprocal altruism, indirect reciprocity, costly signaling, and altruistic punishment. A subfield of economics, game theory, is devoted in part to the empirical testing of these theories with the aim of understanding altruistic behavior. This paper describes the results of a novel game used to test such theoretical predictions. Building on experimental economic games previously performed by Tracer (2003, 2004, Henrich et al. 2001 ), this thesis presents the results of a third-party justice game performed among the Au of Papua New Guinea. In brief, a proposer makes an offer to an inert and anonymous recipient of how to divide an initial endowment. An anonymous third-party is given an endowment equal to 50% of the initial endowment, and the opportunity to sanction the proposer, compensate the recipient, or do both, all at a cost to himself. These costs are a 20% sacrifice to punish, 20% sacrifice to compensate, and a 40% sacrifice to both punish and compensate. Conversely, if the third-party chooses not to act he leaves the game with the full endowment allotted to him. This research design will test the basic economic

PAGE 15

and evolutionary prediction that players will behave selfishly, test the validity of the theory of altruistic punishment as a norm-enforcer, and explore the influence of culture-specifically local ideas about justice-on game behavior. According to the "selfishness axiom," conventional evolutionary and economic theory predicts that in the third-party justice game, the proposer will offer nothing, and that the third-party will never act on any proposal, despite how 'unfair' the offer is to the recipient. However, empirical results from diverse experimental economic games played both among the Au and elsewhere show that in economic games, individuals do not strictly behave as rational, selfish, utility maximizers. Instead, they altruistically offer as much as 50% of their endowment with or without the threat of punishment; individuals also altruistically punish in games (Bolton et al. 1998, Bowles and Gintis 2004, Boyd et al. 2002, Camerer 2003, Cameron 1995, Diekmann 2004, Eckel and Grossman 1998, Fehr and Fischbacher 2004b, Fehr and Gachter 2002, Fehr et al. 2002, Fowler et al. 2004, Henrich and Boyd 2001, Henrich et al. 2001, Hoffman et al. 1996, Nowak et al. 2000, Riechmann 2002, Romp 1997, Roth et al. 1991, Tracer 2003). Moreover, results greatly differ between industrialized and small-scale societies, ostensibly due to cultural mores about fairness and the degree of free-market integration (Henrich et al. 2001 ). Because fairness and justice are inextricably linked, the research design described here also aims to elucidate whether the predominant justice system plays a role in player behavior by allowing the option of punishment, compensation, or both actions at a cost to the third-party. These actions may act as proxies for retributive, restorative, and distributive justice, respectively. 1.2 Justification of the Site Several characteristics make the Au an ideal population for testing game theory. First, results from behavioral economic studies performed in small-scale societies (Henrich 2000, Henrich et al. 2001, Tracer 2003) greatly differ from those 2

PAGE 16

in industrialized nations (Camerer 2003, Cameron 1995, Diekmann 2004, Hoffman et al. 1998, Roth et al. 1991 ). The study of cooperation in smaller-scale settingslike among the forager-horticulturalist Au-may prove more tractable than studies performed in large-scale groups with complex division of labor. Results from small scale societies may provide valuable information about the pre-conditions for cooperation among large-scale groups and thereby help refine theory about the evolution of altruism. Second, typical of Melanesia, the Au have a structured system of reciprocal exchange that influences social, economic, and political interactions, and bolsters social ties within and beyond the family or clan. The enculturation of reciprocal norms may produce interesting (or anomalous) experimental results that may corroborate or detract from previous theoretical assertions about social preferences either based on cultural rules of fairness, or on inequality aversion. Indeed, Tracer (2003) reports that in economic games previously performed among the Au, participants are not only very generous, sometimes offering greater than 50% of the original endowment, they also altruistically punish with high frequency-that is, at a cost to themselves, they punish those who make unfair offers. Further study will help ascertain to what degree this behavior is due to the entrenchment of reciprocal norms. Third, previous game theory experimentation in small-societies and among the Au themselves (Henrich et al. 2001, Tracer 2003, 2004) provide both a foundational guide for the research design reported here and a body of empirical results for comparison. Fourth, game behavior had been said to be largely based on a local sense of fairness (Henrich et al. 2001 ), and justice is implicit in ideas of fairness. Therefore the justice system of a particular society may inform game behavior. In addition to a Western retributive system, Papua New Guinea also utilizes, at least customarily if not formally, a restorative justice system, namely in the emphasis of victim compensation. In such a context, game experimentation that gives the third-party player the option to punish the proposer and to compensate the recipient, both of whom are anonymous, will help 1) determine how ingrained either 3

PAGE 17

type of justice system is, and 2) support the idea that altruistic rewards may be just as important as altruistic punishment for the maintenance of cooperation. 1.3 Specific Aims and Predictions The following hypotheses make predictions about proposer (1.) and third party behavior (11.). These hypotheses flow from the "selfishness axiom" of both economic and evolutionary theories. I. The proposer will never make a non-zero offer. If individuals are selfish, as economic and evolutionary theory predicts, offers in the third-party justice game should not exceed 0%, especially under conditions of anonymity. The frame of the third-party justice game should likewise not induce non-zero offers. The threat of punishment has been the alleged cause of non-zero offers in ultimatum games (See Chapter 2). In that game, a proposer makes an offer to a recipient of how to divide an endowment. The recipient may accept the offer, or he may reject it so that neither player gets any of the endowment. Thus, in the ultimatum game the threat of punishment to the proposer is direct, ostensibly driving proposers to offer non-zero amounts to ensure they leave the game with a positive payoff. In the third-party justice game however, the threat of punishment is indirect; that is, it comes from a third-party, whose endowment is given independent of proposer-recipient action. Because the recipient is inert, and because the threat of punishment is indirect, offers should never exceed the purely selfish, 0%. Cultural norms could, however, drive offers up from the predicted non-zero mark. Among the Au, these influences include (probably a combination of) the ingrained system of reciprocal exchange typical of Melanesian society, which places a premium on generosity; the lack of anonymous interaction in daily life, which would cause participants to believe that their actions in the game might one day 4

PAGE 18

induce detrimental returns should their identity be found out; local ideas about fairness; or even the customary justice system that values victim compensation among other punitive measures. II. Third-party players will never act because it is costly to do so. According to economic and evolutionary theories, the third-party should never punish or compensate because of the 20% cost to do so; that is, it would be an altruistic act. The third-party should especially not choose to both punish and compensate because of the 40% cost to do so. However, empirical evidence suggests that there are exceptions to these assumptions, and that in economic games, individuals sometimes punish other players at a cost to themselves. Such altruistic punishment has previously been rationalized on grounds of fairness; that is, a player responds negatively and punishes a coplayer for what he sees as unfair (here, selfish) behavior (Rabin 1993, Fehr and Gachter 1998). While negative reaction toward unfairness may explain games in which proposer-action directly impacts punisher-utility, in this game the third-party receives his endowment regardless of the proposer's (fair or unfair) behavior. Punishment has also been rationalized on grounds of indirect utility maximization; that is, decreasing the utility of the proposer (and probably the player with the most utility) increases self-utility relative to other players (Tracer 2003). Though both explanations flout standard economic theory, relative utility maximization is congruent with evolutionary theory. Compensation cannot be rationalized on grounds of economic or evolutionary theory. However, like non-selfish offers, compensation may be explained by cultural proclivities that emphasize reciprocity and victim compensation. Punishment and compensation are equally costly, at a sacrifice of20% of the third-party endowment. However, the customary criminal justice system emphasizes victim compensation at 5

PAGE 19

least as part of punition. Unfair offers approximate petty crime more than they do violent crime, which might warrant more stringent punishment. Therefore third-party compensation could be explained, at least in part, by the impact of cultural norms on generosity and justice. However, the concurrent act of both punishing and compensating comes at an enormous cost-a 40% sacrifice of their endowment-so that this behavior should never be observed. 1.4 Significance By adding to a growing body of empirical evidence for the existence of altruism, this thesis will foster debate about the standard predictions of canonical economic and evolutionary theory. Experimental results presented here and elsewhere show that players (and by extension, individuals) do not always behave selfishly. These results suggest there is a need for revision of theoretical generalizations, or at least refinement of theory to help explain how altruism can become stable under expectations of selfishness. Reform of theory will have implications in a range of disciplines including psychology, anthropology, sociology, economics, and political science (Diekmann 2004). The importance of theoretical refinement is not limited to academia, but also bears implications in economic policy and constitution design (Fehr and Gachter 2002, Gowdy and Seidl2004). Standard (Western) economic models based on the selfish, rational, individualistic actor lead to a micro-foundations approach to economic and social policy. This approach assumes perfect competition in capitalistic, impersonal markets that can and will reach equilibrium. In reality, the social nature of humans allows us to learn and base behavior on culture as well as feed-back from other individuals. Acknowledgment of this social nature produces a much less parsimonious yet more accurate model of economic reality, a model with the potential for multiple equilibria. This acknowledgment necessitates a wholesale change of assumptions about consumption. Informed policy based on such changes 6

PAGE 20

may ameliorate growing global inequities produced from the forcing ofneo-liberal economic ideals on developing nations. 1.5 Chapter Summaries Chapter 2 provides an overview of theories invoked to describe the evolution of altruistic cooperation and the use of experimental economic games to test such theory. Following an introduction (Section 2.1 ), the characteristics of altruism are discussed (Section 2.2), including the question of whether or not it exists among non human primates. Section 2.3 details the four main theories for the evolution of altruism, starting with Hamilton's theory of inclusive fitness (Section 2.3.1) and Trivers' theory of reciprocal altruism (Section 2.3.2). These theories rely on a strict interpretation of evolution by individual selection and do well to explain how altruism could arise among kin and among dyads with repeated interactions, respectively. The similar theories of indirect reciprocity and costly signaling (Section 2.3.3) extend the above theories to describe the proliferation of altruism among groups larger than two individuals. Section 2.3.4 addresses the theory of strong reciprocity, or altruistic punishment, to describe the sustaining of cooperation in anonymous interactions; proponents of altruistic punishment use the controversial idea of group selection to explain its inception and evolutionary proliferation. Section 2.4 describes the importance of economic game theory as a burgeoning avenue of social science research into social preferences, beginning with a background and history in Section 2.4.1 and a presentation of assumptions in Section 2.4.2. Section 2.4.3 critically describes and presents empirical evidence of the prisoner's dilemma, dictator, and ultimatum games. Section 2.4.4 briefly enumerates the weaknesses of game theory, and is followed by a theoretical and empirical conclusion in Section 2.5. Chapter 3 presents a brief overview of the study population, the Au of Papua New Guinea, providing description ofthe research setting (Section 3.1) at the 7

PAGE 21

national (Section 3.1.1) and local levels (Section 3.1.2). Section 3.2 is a very brief ethnography of the subject population (Section 3.2.1 ), followed by a description of the reciprocal exchange system that pervades Au society as it does in most Melanesian societies (Section 3.2.2). Section 3.2.3 concludes the chapter with an exploration of law and justice in Papua New Guinea, emphasizing compensatory measures as a manifestation of restorative justice. Chapter 4 provides a more thorough examination of the three main types of justice (Section 4.1 ), retributive (Section 4.1.1 ), restorative ( 4.1.2), and distributive justice ( 4.1.3). The idea that justice and the underlying issue of fairness influence game behavior is explored in the final section ( 4.2), providing a transition into the current research design, which aims not only to explore the altruistic (or selfish) behavior of participants, but also elucidate whether a particular form of justice predominates in game behavior among the Au. Chapter 5 provides details of the research design and methodology. Section 5.1 describes sample recruitment, Section 5.2 explicitly describes the third-party punishment game, and Section 5.3 presents the mechanism of data collection. A summary of the sample population by gender is presented for each of the three villages where the game experiment was performed. The chapter ends with a description of data analysis (Section 5.4) and a brief enumeration of methodological issues (Section 5.5). Chapter 6 consists of the empirical results followed by discussion. Quantitative results make up Section 6. 1, specifically including a presentation of the frequencies of variables (6.1.1), proposer offers (6.1.2), and third-party actions (6.1.3). Qualitative results are summarized in Sections 6.2, with the results of interviews and responses to vignettes (Section 6.2.1 ). Discussion commences in Section 6.3. The final chapter features a conclusion (Section 7.1), final summary of the theories for the evolution of cooperation (7.1.1 ), suggested future directions (7.2), and concluding remarks (7.3). 8

PAGE 22

2. Theoretical and Empirical Background 2.1 Introduction Pure altruism-a costly act performed on behalf of a genetically unrelated individual-poses an evolutionary problem under the assumption that natural selection predisposes all organisms-and arguably the genes they carry-to be selfishly concerned with fitness and reproductive success (Axelrod 1984, Darlington 1972, 1978, Fisher 1992, Hamilton 1964). Altruistic cooperation is only a stable strategy in a population entirely composed of cooperators. Though altruism could arise as a mutation, the proliferation of altruism among a population of egoists is a biological conundrum if altruistic behavior decreases the fitness of altruists while increasing the fitness of beneficiaries (Boyd and Richerson 2005). Theories of inclusive fitness (Hamilton 1964, 1972) and reciprocal altruism (Trivers 1971) attempt to explain how altruism could proliferate in animals including bees, ants, and vampire bats, but they fall short of explaining human cooperation in large groups of unrelated individuals (Henrich and Boyd 2001 ). In fact empirical evidence from experimental economics predicts that natural selection does not favor cooperation for groups larger than ten persons without a high degree of relatedness (Henrich and Boyd 2001 ). Other theories, like those of indirect reciprocity (Fishman 2003, Mohtashemi and Mui 2003, Nowak and Sigmund 1998), costly signaling (Austen Smith and Banks 2002, Gintis et al. 2001, McAndrew 2002, Zahavi 1975, 1977), and strong reciprocity (or altruistic punishment) (Bowles and Gin tis 2002a, Boyd et al. 2002, Fehr and Gachter 2002, Fehr et al. 2003, Fehr and Henrich 2003, Fowler et al. 2004, Gintis 2000a, Gintis et al. 2003) attempt to take up where inclusive fitness and 9

PAGE 23

reciprocal altruism leave off. Ancillary to the debate of the evolution of altruism are questions about the universality of human altruism, or to what degree altruistic behavior takes its influence from social, cultural, economic, and political contexts that produce social norms like fairness. Anecdotal and ethnographic evidence of altruistic human cooperation without obvious external sanctions exists in many forms, from food sharing of hunted game to warfare (Fehr and Gachter, 2002) to giving blood and charitable donations (Boyd and Richerson, 2005). While some acts may be rationalized by the potential for future reciprocity, in others, a 'public good' may confer benefit on an individual who contributes nothing at all. Furthermore, beneficiaries may belong to large anonymous groups with little or distant (genetic or otherwise) relation to benefactors. Benefactors may also act among those with whom they will never interact again (one-shot encounters) so that prestige or reputation gains are minimal or non-existent (Bowles and Gintis 2004, Boyd and Richerson 2005, Fehr and Gachter 2002). In addition to anecdotal accounts, further evidence of human cooperation comes from controlled economic experiments used to test game theories (Berget al. 1995, Bolton et al. n.d., Bolton et al. 1998, Cameron 1995, Clark and Sefton 2001, Diekmann 2004, Eckel and Grossman 1998, Fehr and Fischbacher 2004b, Fehr et al. 2002, Henrich 2000, Henrich et al. 2004, Hoffman et al. 1994, Hoffman et al. 1996, Riechmann 2002, Roth et al. 1991, Telser 1995, Tracer 2003,2004, Wedekind and Milinski 1996). Altruism is a particularly fascinating and arguably universal aspect of human behavior, thereby making it the subject of study in a diverse range of fields from philosophy (Kitcher 1993) to justice, biology, economics, and anthropology. Unlike other universal human activities that have obvious biological functions (e.g., eating, sex), altruism has yet to be fully reconciled in terms of Darwinian natural selection which should disallow its spread and stabilization. This opens the door for the alternative explanations, proposed by sociobiologists (Fisher 1992, Wilson 1983) and 10

PAGE 24

selfish-gene theorists (Dawkins 1976), of group and multi-level selection (Alexander 2000). Before proceeding to theories for the evolution of altruistic cooperation, issues in defining altruism will be discussed. 2.2 Human and Non-human Altruism A number of messy definitions fall under the heading of altruism. Most generally, any act costly to one's self and beneficial to another is altruistic. But this would include such basic and instinctual behavior like the care of offspring (Fisher 1992). Others definitions eliminate nepotistic acts of kindness or care for descendants and relatives from the repertoire of altruistic endeavors, to define altruism as a costly act on behalf of a genetically unrelated individual. A further issue is whether to include reciprocity under the heading of altruism. If an altruistic act is committed with the expectation of return then perhaps the contingent exchange is better defined as reciprocal mutualism. However, Darlington (1978) argues that any altruistic act is potentially reciprocal: a beneficiary may decide at a later date to return a favor. This further complicates the dividing line between altruism and reciprocity necessitating one further distinction on grounds of intent: an altruistic act is one that is made without expectation of return. Pure altruism is very rare or arguably non-existent among nonhuman animals, though as mentioned before, various forms of cooperation are often lumped together under altruistic headings including kin-related cooperation and mutualism. Axelrod ( 1984) argues that though selfishness is the primeval, evolutionarily stable strategy1 cooperation may evolve when individuals interact repeatedly and frequently. This is most often enacted by related individuals, but may also arise among higher 1 Whereas the Nash equilibrium describes the best strategy in a given environment in economics, the equivalent in behavioral anthropology is the Evolutionarily Stable Strategy (ESS), or the schema that cannot be undermined in a given ecological niche unless something in the environment changes (Maynard-Smith 1982). II

PAGE 25

organisms with sufficient cognitive ability to both recognize previous interactive partners and to remember results of previous interactions so that the appropriate response may be chosen. While Axelrod argues that mutualism is a primitive example of altruism, others (Brosnan and de Waal 2002) contend that, because mutualism confers spontaneous benefit on the actors, it is a category all its own, separate from altruistic and reciprocal behavior typical of gregarious animals like humans and other non-human primates (Silk 2004, Terborgh and Janson 1986), among whom group living and constant, repeated social interactions set the stage for cooperation. There are many examples of cooperation from the non-human primate world. These acts may or may not be altruistic, depending on the relatedness of group members. Chimpanzees stage collective border patrols and seemingly organized attacks resembling warfare on neighboring groups (Alexander 1979, Silk 2004); vervet monkeys sound alarm calls when predators encroach and may come to the aid of allies and kin when in danger (Alexander 1979, Seyfarth et al. 1984); captive and wild cebus monkeys, capuchins, and chimpanzees share food (Brosnan and de Waal 2002, Conroy 2005, McGrew 1998, Silk 2004); macaque females defend unrelated juveniles from harassment by other group members (Chapais et al. 2001); Iangurs and howlers carry other females' infants (Silk 2004); and most primates spend 1020% of their waking hours grooming group members. Primates also sometimes engage in polyspecific cooperative behavior; for example, squirrel monkeys form alliances with capuchins, and both species seek out associations with tamarins (Terborgh and Janson 1986). Though inconclusive, it seems that some of the same advantages to intraspecific group living may apply to groups of different species (Reed and Bidner 2004), namely predator avoidance and increased knowledge about, access to, and defensibility of food resources (Fleagle I 999, Terborgh and Janson 1986). 12

PAGE 26

Among group-living species, Axelrod ( 1984) proposes the following potential keys to cooperative tendencies: 1) complex memory, 2) information processing that uses past effects to inform present action, 3) ability to predict the probability of future interactions (with the same individual), and 4) enhanced ability to distinguish between different individuals and their associated behaviors. The latter may in fact be the most important attribute, allowing individuals to favor some and punish others, depending on previous behavior, and therefore protect oneself from repeated harm from defectors. These qualities are naturally difficult to assess and are thus controversial among nonhuman primates. The above primate social activities may be considered to be altruistic to different degrees. With the exception of polyspecific cooperation, it seems that most of these activities demonstrate cooperative behavior between genetically-related individuals, so that these acts are not examples of pure altruism. Moreover, Brosnan and de Waal (2002) assert that low-cost, opportunistic reciprocity should not be categorized with high-cost, altruistic reciprocity. In other words, it is much less risky to sometimes participate in low-cost allo-grooming than to risk one's life in the protection of conspecifics from predators. The above primate activities might be better categorized as cooperation based on high rates of association or mutual tolerance, not on contingent exchange. This distinguishes primate behavior from human cooperation that is generally based on score-keeping, or remembering previous interactions (Brosnan and de Waal2002, Silk 2004). Humans not only cooperate with kin and those with whom they have repeated interactions, but with strangers as well. Both seem to violate the selfishness axiom of evolutionary and economic theory because though individuals can receive benefit from mutual cooperation with other individuals, they can benefit even further by exploiting the cooperative actions of others (Axelrod 1984). Defection in one-shot, random encounters makes perfect evolutionary sense, if heritable will create a population of defectors, and is an evolutionarily stable strategy. But defection 13

PAGE 27

becomes detrimental when interactions are repeated and previous defection, if remembered, can prompt punitive measures. If a cheater should meet again those he has exploited, the cheated may punish the cheater. In summary, there is a contentious continuum of altruism (Humphrey 1997), with pure altruism at one end--especially that in one-shot, anonymous encountersand kin-related care (e.g., parent-offspring care) at the other. Mutualism, reciprocity, and cooperation lie somewhere in between, depending on the degree of relatedness of actors, delay in receiving returns, and cost of the altruistic act. A growing body of empirical evidence from experimental economics shows that non-selfish behavioral preferences are not anomalous (Axelrod and Hamilton 1981, Camerer 2003, Cameron 1995, Clark and Sefton 2001, Diekmann 2004, Eckel and Grossman 1998, Fehr and Gachter 2000a, 2000b, 2002, Henrich 2000, Henrich et al. 2001, Nowak et al. 2000, Roth et al. 1991, Tracer 2003). Altruistic behavior is not confined to individuals, but may confer benefit to a group; thus it is sometimes called 'prosocial' behavior (Henrich 2001, 2004, Gowdy and Seidl 2004). In recent years economists, dissatisfied with inaccurate predictions from neoclassical economic theory, have turned to anthropologists and evolutionary biologists for theoretical explanations of observed behavior. Specifically, these economists seek answers as to how cooperation is maintained among large groups of unrelated individuals and in one-shot encounters (Henrich et al. 2004). The following is an exploration of the various theories for the evolution of altruistic cooperation. 2.3 Theories for the Evolution of Altruism 2.3.1 Kin Selection In 1964, Hamilton proposed a 'genetical mathematical' model for the evolution of social behavior, specifically behavior difficult to reconcile under the principle of classical natural selection. His model follows the same general mechanism of natural selection but is based on a refinement of the concept of fitness, 14

PAGE 28

to encompass not just the fitness of an individual but that of an individual plus all of his genetically related kin. The unit of Darwinian natural selection is the individual, where fitness is measured by the relative number of that individual's offspring that survive to reproduce. Dawkins ( 1976) argues that findings in genetics have moved focus from the individual to the heritable gene as unit of selection. It is gene preservation and proliferation that is at the core of individual survival. This tweaking of the definition of fitness bolsters the assumptions at the core of Hamilton's theory of kin selection. When the unit of selection is not the individual but his genes, fitness and reproductive success take on new nuances because an individual shares common genes with his kin (Fisher 1992). Reproductive success then, becomes the additive reproductive viability ofboth vertical (descendent) and collateral (non-offspring) relatives (Alexander 1979, Alexander and Hamilton 1981, Daly and Wilson 1978, Fisher 1992, Hamilton 1975). For the positive selection of a gene, an increase in individual fitness is not enough if it comes at the expense of related individuals who may carry replicas of the same gene (Hamilton 1962). For example, siblings share 50% (on average) of their genes so that protecting and helping a sibling protects 50% of one's own genes2 (Daly and Wilson 1978, Dawkins 1976, Fisher 1992). Conversely, a gene disadvantageous to an individual may undergo positive selection if that gene confers advantage on relatives. Therefore, if an individual who has the genes for altruism helps a relative, this gene is shared in the beneficiary relative, and the beneficiary survives and reproduces, the altruistic gene will also be replicated (Fisher 1992). Inclusive fitness is a quantity incorporating the overall number of genes of related individuals, so that a positive net increase in this quantity will null the negative fitness cost to the altruistic individual (Alexander 1979, Chapais 2001, 2 In other words, it makes as much sense to feed one's siblings as it does to feed one's children (Daly and Wilson 1978). 15

PAGE 29

Chapais et al. 2001, Darlington 1978, Fehr and Gachter 2002, Hamilton 1964, 1975, Rodman 1999, Silk 2004, Strier 2000). If altruism is the bestowal of reproductive benefit (b) upon a beneficiary at some cost (c) to the benefactor ( Chapais 200 1, Chapais et al. 2001, Daly and Wilson 1978, Strier 2000), the altruistic act may be selectively advantageous if the beneficiary and benefactor carry an allele in common. The probability (r) of carrying that allele is the coefficient of relatedness of the two individuals3 For the allele to be positively selected, the cost of the beneficial act must be less than the benefit multiplied by the probability of relatedness ( c
PAGE 30

sterile worker bee cannot reproduce, but altruistically helps its fertile relative, thereby making it appear that it is the family that is the unit of selection, when it is really the gene, as shared by family members. Inclusive fitness-driven behavior is phenotypically altruistic, but genotypically selfish (Alexander 1979) where inclusive fitness may be maximized through the favoring of more closely related individuals. This is different from the controversial theory of group selection. Though altruistic acts among related individuals could be termed group selection ofthe genes those individuals carry, it is not selection for individuals, because those individuals are not wholly identical. Hamilton's theory of inclusive fitness is a feasible model for the proliferation of altruistic cooperation in small groups of genetically-related individuals (Boyd and Richerson 2005). However, some argue that Hamilton's formula for kin selection is far too simplistic-mathematically as well as in the assumption that altruism could be heritable via a single locus allele-and that it exaggerates the probable effectiveness of kin selection (Darlington 1978, Matessi and Karlin 1984). A further problem is that the theory seems to imply that individuals are endowed with a great deal of rationality in choosing to help relatives, presumably with the (unconscious, yet) ulterior motive of bestowing kind acts on those from whom future (genetic) returns are most likely. In addition, inclusive fitness fails to explain altruistic cooperation when group size increases or the relatedness of group members decreases. Furthermore, humans continue to cooperate even in one-shot encounters with strangers (Boyd and Richerson 2005, Henrich et al. 2004), and when cooperation transcends the boundary between species (Fisher 1992). 2.3.2 Reciprocal Altruism After inclusive fitness, the second solution to prosociality is the idea of reciprocal altruism. Championed by Trivers ( 1971 ), reciprocal altruism is the manifestation of an unconscious heritable trait that prescribes cooperation in dyadic, 17

PAGE 31

long-term, repeated interactions (Axelrod 1984, Boyd and Richerson 2005, Bradley 1999, Fehr and Gachter 2002, Fisher 1992, Henrich et al. 2004, Silk 2004). In contingent exchanges, two or more individuals, regardless of kinship, profit more than the cost of individual altruism (Darlington 1978, Diekmann 2004, Fishman 2003). Returns may arrive at any time after the initial act (Fishman 2003). Reciprocity and altruism differ in that 1) reciprocity is conditional fairness, as opposed to the unconditional generosity of altruism; 2) reciprocity is not categorical (i.e., all or none) but is dimensional, so that it is possible to reciprocate to a degree; 3) reciprocity is an obligation evoked by previous behavior, but returns may be of a different currency than that in the original exchange (i.e., tit-for-tat= eye-for-an-ear [or something else] or tat-for-tat= eye-for-an-eye); 4) the reciprocal norm does not only apply to benevolent action; punishment may be the return; 5) reciprocity may be driven by egotistic motivation, to get something in return (from material goods to heavenly honor, which may explain seemingly altruistic charitable donations) (Gouldner 1960). Trivers' (1971) quintessential example is that of cleaner-fish, or the small fish that trades cleaning of the mouth, teeth, and gills of a larger fish in exchange for a meal (and not being devoured). Extending this model of contingent exchange to humans, Trivers used game theoretic methods to argue that the reciprocal strategy is willing cooperation in the first interaction followed by conditional, continued cooperation based on previous partner action; this strategy is also known as tit-for-tat (TFT) (Axelrod 1984, Bartholdi et al. 1986)4 In reciprocal altruism, cooperation is responsive and based on compatibility more than on kinship (Clark and Sefton 2001). Moreover, ifthere is a gene for reciprocal altruism, survival 4 In computer tournaments, TFT never defects first, but always defects immediately after an opponent defects, and always cooperates immediately after an opponent cooperates (Bartholdi et al. 1986). 18

PAGE 32

will be enhanced by natural selection acting on mutual benefits5 This is because cheaters6 will be selected against if the act of cheating has future costs (i.e., being shunned from future cooperation) that outweigh the immediate benefit of cheating (Trivers 1971 ). Dispersal strategies to proscribe inbreeding require many animals to live in groups with unrelated individuals. The theory of reciprocal altruism posits that unrelated individuals may cooperate with the expectation of returns at a later time when interactions are repeated often enough for previous partners to be recognized (Henrich et al. 2004, Strier 2000, Trivers 1971 ). Several pre-conditions enhance the likelihood of success of reciprocal altruism, including a long lifetime (so that reciprocal encounters may be frequently experienced and remembered), low dispersal rates, high mutual interdependence (including extended parental care), and the existence of stable social groups (Trivers 1971 ). While Trivers (1971) contends that cleaning fish and alarm calls in birds are examples of reciprocity altruism, others assert that animals are incapable of pure reciprocity among unrelated individuals (Alexander 1979, Bradley 1999). Alleged reciprocal evidence from the non-human primate world includes grooming among vervet monkeys, where grooming between unrelated individuals increases the chances that an individual will respond to alarm calls (Seyfarth and Cheney 1984); and the sharing of food among chimpanzees, ostensibly based on past interactions (Fehr and Gachter 2002), with some evidence suggesting that chimps do so in memory-based, partner-specific exchanges (Brosnan 5 Hamilton would argue, however, that by virtue of obtaining a return for an altruist act, the act is no longer altruistic by definition (Humphrey 1997). It is simple reciprocity, or an 'eye for an eye.' Trivers would retort that in inclusive fitness, individuals do altruistic acts on behalf of genetically related individuals and therefore is not pure altruism by definition. Humphrey ( 1997) argues however, that genetic returns of inclusive fitness are returns so that kin selection has an element of reciprocity to it. The difference between the two then, is intent: the kin altruist most likely does not expect a return, so that reciprocal altruism is not altruism because there is an expectation of return. 6 Following Trivers ( 1971 ), a 'cheater' is an individual who fails to reciprocate, without referring to intent or intending to imply that the individual is morally defective. 19

PAGE 33

and de Waal2002) so that sharing is more likely among chimps who have shared food in the past. For example, if chimp A requests food from B, B will be more likely to respond with aggression if chimp A has not shared with B in the past (de W aal 1991 ). Some argue that this behavior in chimps proves that reciprocation is buried deep in human evolution (Fehr and Gachter 1998). Though the costs and benefits incurred from acts of proposed reciprocal altruism are difficult to measure (Brosnan and de Waal2002, Chapais et al. 2001, Strier 2000), theory predicts that cooperation among unrelated animals can arise when they interact regularly and have the opportunity to adjust their (cooperative) behavior according to previous experiences (Seyfarth and Cheney 1984). The potential for these interactions occurs in various animal societies typified by group living, as shown in the examples above. However, it does seem that overwhelmingly, most reciprocal encounters in the animal kingdom occur between genetically related individuals, while among humans that is not the case. Reciprocity may best be summed by the adage 'an eye for an eye,' encompassing both the concepts of positive and negative reciprocity (Fehr and Gachter 1998). Equal returns for equal gains underlies both concepts, yet negative reciprocity may be more aptly viewed as punishment or revenge while positive reciprocity is driven by the desire to return kindness to those who have been previously kind (Rabin 1993). Two problems remain: how can reciprocity arise under conditions of natural selection if it does not begin among kin; and how are cheaters recognized (Fisher 1992)? The second problem increases in importance as group size increases, especially when accurate recognition of previous cooperators is required (Henrich et al. 2004). 2.3.3 Indirect Reciprocity and Costly Signaling Trivers ( 1971) defines two distinctions in reciprocity: 1) direct reciprocity results when rewards come from the same individual who receives original benefit; 20

PAGE 34

2) indirect reciprocity results when rewards come back from the cooperative social group at large in the form of prestige or respect (Alexander 1979, Fishman 2003, Mohtashemi and Mui 2003, Nowak and Sigmund 1998). In other words, though Mr. X did not do a favor directly for me, he did so for Mr. Y and for Mrs. Z. I can then infer Mr. X is likely to return a favor to me. While direct reciprocity requires repeated dyadic interactions, indirect reciprocity may facilitate cooperation in larger groups so long as information about individuals' reputations may continually be assessed. Some argue that while non-human primates may be capable of direct reciprocity, indirect reciprocity is unique to humans (Alexander 1979). Proceeding from the field of psychology, costly signaling is akin to indirect reciprocity. In costly signaling, actions or objects are used to communicate some non-obvious information about intention or personality to other players (Austen Smith and Banks 2002; Gin tis et al. 200 I; McAndrew 2002). Signaling has been invoked to explain seemingly inefficient market activities like advertising and factory strikes, and has been used in biology to explain morphological traits such as peacock feathers and apparently maladaptive traits like heavy, large elk antlers. Other examples include product warranties, college degrees from a prestigious university (Camerer 2003), and large corporation or celebrity donations leaked to the media7 Costly signaling is sometimes called competitive altruism because an individual may perform acts of extreme philanthropy in times of plenty in order to position themselves for access to resources in later times of need (McAndrew 2002). In this way, costly signaling is a good strategy for inducing reciprocal altruism. But costly signaling may apply to a range of social interactions like resource sharing, 7 For example, during the writing of this thesis, the nation of Kuwait issued a public 'costly signal' in the devastating aftermath of Hurricane Katrina (New York Times 2005). Dr. Anas Al-Rasheed, Kuwait's Minister of Information, wrote a full-page ad in the New York Times to express his country's condolences to the victims of the hurricane and to pledge $500 million in aid for the reconstruction ofthe Gulf Coast. Though he worded the pledge as a reciprocal offering in thanks for U.S. ''help" during the invasion of his country by Iraq and in the subsequent rebuilding of Kuwait, the very public pledge doubled as a costly signal of Kuwait's political alliance and economic status. 21

PAGE 35

defense, raiding, and the punishing of free-riders or norm-violators (Gintis et al. 2001). The 'quality' of the signaler is the genetic or phenotypic attribute that is difficult for others to assess directly. Yet, it has effects on payoffs from social interactions with the signaler: those who give benefits or give more benefits (those who signal more intensely) are advertising their good qualities. This influences future behavior of group members who might allocate payoffs to the signaler, for example, choosing them as allies or mates, or choosing to defer to them in competition. Costly signaling may also benefit the group by virtue of sharing social information (McAndrew 2002). Signals are an evolutionarily stable, sustaining force of cooperation if they are less costly than the benefits accrued when others 'decode' the signal and reciprocate; and are too costly for cheaters to fake (Camerer 2003, Henrich et al. 2004). Four qualities of a costly signal are that 1) the behavior must be easily observable, 2) must be costly to the signaler (in resources, energy, etc.), 3) must be an honest indicator of some real trait (otherwise it is just 'cheap talk ')(Austen-Smith and Banks 2002), and 4) must provide some advantage to the signaler (McAndrew 2002). In order to be altruistic, costly signals must also lead to some advantage to the observer. Indirect reciprocity and costly signaling may maintain cooperation in groups where cooperators build a reputation and do not necessarily obtain material returns (Fehr and Gachter 2002). The theory of costly signaling asserts that cooperation evolves because it is a signal of the group member's quality as a mate, ally, or competitor, and later facilitates the formation of advantageous alliances for signalers (Gintis et al. 2001). Both indirect reciprocity and costly signaling add the important element of addressing altruistic evolution among multiple members, not just between dyads. Though indirect reciprocity and costly signaling go farther in the explanation of cooperation among unrelated groups of people than do inclusive fitness and 22

PAGE 36

reciprocal altruism, these theories do nothing to ameliorate the problem of one-shot or anonymous cooperation where there is no chance for signaling and no chance to observe previously cooperative behavior (Bowles and Gintis 2002a, Henrich et al. 2004); nor do they explain cooperation when groups are very large and assessment of individual reputation is difficult or impossible (Diekmann 2004, Mohtashemi and Mui 2003). 2.3.4 Altruistic Punishment Cheaters who take advantage of a purely reciprocal system threaten to destroy reciprocal cooperation. Punishing these cheaters, however, reinforces the altruistic system, and has been called negative reciprocity (Fehr and Gachter 1998), strong reciprocity (Gintis 2000a, Gintis et al. 2003), or altruistic punishment (Fehr and Gachter 2002). Strong reciprocators are cooperators who punish non-cooperators even if punishing is costly, even in anonymous and one-shot circumstances, and even if the cost of punishment is unlikely to be repaid (Bowles and Gintis 2002b, Gintis 2000a, Gin tis et al. 2003, Henrich et al. 2001 ). If this form of interaction is termed strong reciprocity, then the reciprocal altruism championed by Trivers ( 1971) is weak reciprocity (Gowdy and Seid12004). Some (Bowles and Gintis 2004, Boyd et al. 2004, Gintis 2000a) argue that strong reciprocators are more likely to survive when their group faces extinction than either reciprocal altruists (who rely on repeated interactions) or costly signalers (indirect reciprocators)8 as punishment could maintain norms of cooperation better than these under duress. Self-interest with a view to future success is at the core of both reciprocal altruism and inclusive fitness, where self-interestedness allows natural selection to favor these mechanisms of 8 Arguably, altruistic punishment may confer a fitness advantage on social-groups with an above average number of punishers so that they may better survive group-threatening catastrophe, like war and famine (Sigmund eta!. 2002). Under these conditions, cooperation breaks down if individuals behave selfishly; punishers thereby discipline the selfish for the good of the group. 23

PAGE 37

cooperation in repeated interactions (Boyd and Richerson 2005, Gintis et al. 2003). Alternatively, altruistic punishment, or strong reciprocity, is built on the principle of reciprocity as discussed above, but is more difficult to explain in terms of self interest (Gintis et al. 2003). Strong reciprocity is a form of altruism because it benefits the group at large while the punisher suffers costs. Both ethnographic and empirical evidence show that people punish defectors not just in repeated interactions, but also in one-shot encounters (Boyd and Richerson 2005, Fehr and Gachter 2002). Strong reciprocators inflict various forms of retribution, including refusal of future cooperation; physical attack; social ostracization through rumors, gossip, or imprisonment; and denial to resources, territory, or mates (Boyd and Richerson 2005). Results from ultimatum games (See Section 2.4.3) and public goods experiments9 show that when costly punishment is allowed in game theory experiments, cooperation does not decline; and in games of anonymous pairs, cooperation actually increases when punishment is allowed (Bowles and Gintis 2002a, 2002b, Fehr and Gachter 1998, 2002). In a repeated public goods experiment, Fehr and Gachter (2002) found that the threat of punishment deterred defection while punishment itself acted to reform defectors in subsequent rounds, thereby increasing the prevalence of cooperation (Fehr and Gachter 2002). Bowles and Gin tis (2002b) argue that the motivation to punish is stronger when the identification of the group is known (even if individuals are anonymous), so that strong reciprocity becomes stronger when group stability is high. Thus retributive sanctions may explain why cooperation can be evolutionarily stable among large groups of unrelated individuals (Boyd and Richerson 2005). Building on Gintis (2000), Gintis et al. (2003) assert that strong reciprocity itself is an evolutionarily stable strategy and that even a small fraction of strong 9 Participants are given an endowment and may contribute any amount to a 'group project,' earning a percentage of the total project fund. Defecting, or keeping all the money, seems to be the most profitable strategy, but when all members donate to the group project, joint gains are maximized. 24

PAGE 38

reciprocators could invade a group dominated by egoists and proliferate. The authors acknowledge the explanatory importance of inclusive fitness and reciprocal altruism for the inception of altruistic cooperation, but assert that once structured social interactions are established, strong reciprocity prevents defection, or the pilfering of benefits from the cooperating majority (Boyd and Richerson 2005). Thus strong reciprocity may further explain the evolutionary success of the human species via large-scale and anonymous cooperation, and may be an example of gene-culture coevolution via multi-level selection. Retribution in altruistic punishment is costly to the punisher but beneficial to the group. Over time, punishing the defectors and reforming them towards cooperation not only deters the defecting individual and protects the group as a whole from that defector, it also threatens other potential defectors and deters them from bad behavior (again, threat deters defection, and punishment enhances cooperation). So, long-term benefits outweigh the short-term costs of punishment. When punishers are common, defectors are selected against because they are punished. Selection thus favors punishment, though the resulting cooperation may not equalize the costs to the individual punisher. Altruistic punishment is interesting because I) cooperation may be possible in much larger groups than with mere 'non-policed' reciprocity; and 2) punishers pay the cost to essentially provide a public service for the good of the group. Because of this cost however, selection should favor cooperating non-punishers: individuals who cooperate in exchanges, but who do not incur costs for policing the behavior of non-cooperators. These second order free-riders may exploit punishers if the cost of punishment is high (Henrich and Boyd 2001, Henrich et al. 2004) so that they get higher payoffs than the punishers do. Though the existence of second order free riding shakes the evolutionary stability of punishment (Henrich et al. 2004), 25

PAGE 39

... the payoff disadvantage of altruistic cooperators relative to defectors is independent of the frequency of defectors in the populations, whereas the cost disadvantage for those engaged in altruistic punishment declines as defectors become rare because acts of punishment become very infrequent. Thus when altruistic punishers are common, individual level selection operating against them is weak. [Boyd et al. 2003: 3531] In other words, altruistic punishment becomes stable when punishers are common, thereby effectively limiting the number of defectors and thus the need to punish. This in tum limits the impact of second-order free-riding. Cooperators have a higher fitness than defectors if punishers are common enough so that the cost of being punished is greater than the cost of cooperating. Thus, everyone is better off when punishment exists, yet there are no incentives for any one individual to punish (Fehr and Gachter 2002). Punishment of free-riders then becomes a second-order public good, benefiting group members in the future if they amend their behavior. Proponents of altruistic punishment (Boyd et al. 2003, Henrich and Boyd 2001, Henrich et al. 2004) assert that a cultural evolutionary model based on group selection helps explain the evolution of punishment among humans (Boyd and Richerson 2005, Gintis and Bowles 2003, Sigmund et al. 2002). At the core of the model is the assertion that humans possess a unique form of social learning, namely the copy-cat disposition that allows learning primarily through two innovative mechanisms: 1) conformist transmission, or the copying of frequently observed behavior acted out by the majority; and 2) payoff biased transmission10, or the copying of particularly successful behavior. Supported by empirical evidence from psychology, these types of social learning allow humans to essentially leapfrog other animals-that must learn behavior through parental observation and experiential trial-and-error-into more efficient and adaptive behaviors. 3) The third mechanism, punishment, acts as a behavior regulator to discipline non-conformers (Henrich et al. 10 Also called prestige-biased transmission (Henrich et al. 2004 ). 26

PAGE 40

2004), while the fourth and final mechanism, 4) normative conformity-the drive to match behavior to that of peers-then arises so that despite personal intention or beliefs about the worth of the behavior (and therefore contrary to conformist transmission) the behavioral majority rules. Altogether or individually, the four mechanisms-bimodal cultural transmission, punishment, and normative conformity-help create and maintain behavioral equilibria (like cooperation) not allowed in strictly genetic evolutionary processes. These behavioral equilibria also allow group selection to attain greater importance11'12. Once common, cultural group selection may facilitate the spread of cooperative behavior to non-cooperative groups as members of these groups 1) observe the higher productivity of cooperative groups in either resource allocation or in militaristic domination; and as they 2) imitate, in increasing frequency, individuals with higher payoffs that belong to cooperative groups. Once cooperation and punishment are established through this cultural evolutionary process, prosocial genes can then proliferate because they will increase fitness by making individuals less likely to befall punishment. In this genetic cultural co-evolutionary, multi-level process, rapid cultural changes within-groups can drive group equilibria, where they may remain stable until between-group selection favors an alternate strategy. Henrich (2004) argues that between-group selection will favor prosocial groups because they out-compete groups predominated 11 The authors (Henrich and Boyd 200 I, Henrich et al. 2004) argue that group selection, the whipping boy of many anti-sociobiology anthropologists and geneticists, is not an entirely separate process from individual or natural selection and is a useful concept in order to emphasize the interaction of genes and environment, especially when addressing cultural evolution. Per their definition, 'genetic group selection' comes about when natural selection acts on differences in gene frequencies between groups and overtakes within-group forces, to select for a different equilibrium than that selected by within group forces acting on individuals alone. Cultural group selection works in the same way on learned behavior. 12 An alternate view is that the concept of group selection unites the previously disparate theories of inclusive fitness, reciprocity, and game theory (Wilson 1983) by helping explain (anomalous) behavior that limits each theory alone. 27

PAGE 41

by non-cooperators. When this occurs and cultural selection favors prosocial phenotypes, selection of the prosocial genotype becomes possible. The unique ability to internalize culture described above permits humans to create and follow social norms. When norms are subverted or ignored, negative emotions-resentment, anger-drive those who feel wronged to take action, often sanctioning the norm-subverter (Bowles and Gintis 2002a, 2002b, Boyd et al. 2003, Fehr and Fischbacher 2002a, 2002b, 2004a, 2004b, Fehr et al. 2002, Fehr and Gachter 2002, Gintis 2000a, Henrich and Boyd 2001). These negative emotions allegedly drive punishment in both repeated and one-shot interactions, as strong reciprocity acts as a powerful norm enforcer. Moreover, punishment ostensibly produces shame in the defector because the punished tend to cooperate after punishment13 Fehr and Gachter (2002) emphasize the emotionality of this process, where punishment is a vengeful end, not a rational means toward reform on behalf of public well-being. In sum, altruistic punishment, or the costly punishment of defectors without any material gain, is driven by the proximate mechanism of negative emotion toward defectors (Fehr and Gachter 2002). Both laboratory-controlled experiments and ethnographic data show that players punish defectors not just in repeated interactions, but also in one-shot encounters (Boyd and Richerson 2005, Fehr and Gachter 1998, 2002, Fehr et al. 2002, Henrich 2000, Henrich et al. 2001, Henrich et al. 2004, Tracer 2003, 2004). Including punishment in the explanation of large-scale and one-shot cooperation helps to fill the gaps in the theories of reciprocal altruism, inclusive fitness, and indirect reciprocity (Bowles and Gintis 2002a). Though altruistic punishment does well to explain one-shot anonymous behavior in games, is it a plausible theory for reality? Though modem military, legal, 13 I contend however, that prior defectors may not be reformed toward cooperation because of shame, but because the costs of continuing to defect are too great when defection will probably be punished. Thus future cooperation after punishment may again be explained in terms of self-interest. 28

PAGE 42

and political situations may be isolated, one-shot events (Holt and Roth 1997), how often did our ancestors interact under utterly anonymous and non-repeated conditions? Can results from sterile laboratory experiments be extrapolated to realistic conditions? With the exception of inclusive fitness, the argument could be made that the other theories for the evolution of altruistic cooperation explicate stable conditions once established without offering up explanatory mechanisms for the viability of such strategies in environments dominated by egoists (Axelrod and Hamilton 1981 ). It seems most likely however, that prosociality among humans is far too complex to warrant a simple, parsimonious theoretical explanation. It is feasible that natural selection has acted on environmentally-contingent stable strategies throughout our evolution, including some or all of the above theories (Henrich et at. 2004). This complicated history of altruistic evolution is compounded by the fact that altruistic behavior is extremely difficult to assess, especially among nonhuman animals because of lack of observational data (Brosnan and de Waal 2002, Chapais et al. 2001). Other issues include 1) the difficulty of assessing costs and benefits, especially when the two are of different currency or relative value to different individuals based on age, size, or rank; 2) the confounding effects of reciprocal altruism and kin selection, that is, the difficulty in assessing genetic relatedness of individuals; and 3) temporal issues that make it difficult to determine whether and when returns are made. To bolster debate about the evolution and maintenance of altruism, further empirical testing in varied environments is needed. In recent decades, experimental methods devised in the field of economics have been employed to test whether or not humans are innately selfish utility maximizers, with the aim of further understanding human altruism. 29

PAGE 43

2.4 Behavioral Economics 2.4. 1 A Description and History According to Camerer ( 1999, 2003), behavioral or experimental economics is a reunification (for they used to be more closely cooperative) of psychology and economics in order to explain human behavior. Understanding the cognitive processes that drive humans informs and augments the mathematical rigor of economics. Proceeding from a standard economic approach, and based on Nash's idea that there exist equilibria of optimal strategies, behavioral economics and game theory transcend the limited principles of human rationality that dominate economic theory toward alternative and more realistic theories of behavior offered by psychology. While analytical game theory is a mathematical modeling of what players with varied cognitive abilities will do in certain situations (the game), behavioral game theory is concerned with what players actually do in games; it thus involves more qualitative concerns such as actors' intentions and emotions. Experimental economic games test the predictions of game theory that individuals are innately self-interested in a controlled environment (Romp 1997). Biologists, philosophers, political scientists, and social scientists use game theory to test theoretical predictions about other social preferences and tastes that influence human behavior (Camerer 2003, Gintis 2000a, Gowdy and Seidl 2004, Holt and Roth 2004, Kreps and Rubinstein 1997). Shapley (1953) defines a 'game' as a set of rules that govern the possible actions ofplayers: participants are presented with the game's rules and then strategize accordingly. Person A anticipates what Person B will do; Person A also anticipates what other players will think of his own actions. In repeated games, players may base strategies on previous interactions so that results are different than in one-shot games. Payoffs represent utility, the economic proxy for biological fitness, so that stability of choice depends on the relative fitness (utility) outcome of a game's strategies (Fishman 2003). Results from game experimentation produce a 30

PAGE 44

sort of mathematical x-ray of social situations and the decisions that drive them (Camerer 2003). Game theory has its roots in the work of von Neumann and Morgenstern in the 1940s (Holt and Roth 2004, Kreps and Rubinstein 1997). After World War II, formal game theory rapidly expanded, culminating in a number of advances. The most important of these is probably the Nash equilibrium, an extraordinary one-page proposition published in 1950 that earned Nash the Nobel Prize some 44 years later (Camerer 2003, Holt and Roth 2004, Kreps and Rubinstein 1997). In this paper, submitted while still a graduate student, Nash proposed the solution to how rational players will behave in n-person games when players have a finite set of strategies, including 'mixed strategies,' 14 that correspond to a payoff to each player. Nash's theory predicts that after adjusting their strategies, players will reach a stable equilibrium wherein no player can benefit from a change in strategy. In other words, the theory predicts that each player will choose the best, utility-maximizing strategy in anticipation of other players' strategies (Camerer 2003). Should all players announce their strategies simultaneously, none would want to change his choice (Holt and Roth 2004). With the Nash equilibrium at its core, from the 1970s on game theory gained recognition and moved from its esoteric roots to join the mainstream language of economics (Kreps and Rubinstein 1997). Game theory is given the credit for bringing experimental methods into the field of economics as it lays out testable theoretical predictions about strategic behavior (Holt and Roth 2004). Experimental economists put real people in controlled laboratory settings, to observe their behavior when playing for real (usually monetary) payoffs. Game theory is powerful because of its generality and mathematical precision. Games are usually repeated to understand the proclivity toward equilibrium. There is a growing body of literature in experimental economics 14 'Mixed strategies' are probability distributions over decisions; that is, a player may sometimes choose different actions, for example a poker player who sometimes bluffs (Holt and Roth 2004). 31

PAGE 45

both supporting and contrary to equilibrium predictions. The predictive assumptions of the Nash equilibrium are useful both when accurate and when they do not predict game behavior, because they highlight the existence ofprosocial proclivities. 2.4.2 Homo oeconomicus Economists define pure altruism as the act of increasing another individual's utility at a cost to oneself (Camerer 2003). A fundamental assumption in game theory stems from two mainstays of canonical economic theory, that players (i.e., humans) are both 1) highly rational and 2) inherently self-interested utility maximizers. While the latter echoes the selfishness axiom of standard evolutionary theory, the former endows individuals with more self-awareness and individual autonomy than evolutionary theory assumes. The rational axiom of economic theory assumes that individuals will evaluate every strategy available to them in reality, and by extension, in experimental games. After such evaluation, players in a game will choose the strategy yielding the most desirable-or utility maximizing-strategy according to self-interested tastes and preferences (i.e., they will tend toward the Nash equilibrium, or subgame perfect equilibrium that exists when there are multiple strategies available) (Henrich 2000, Nash 1950a, Shapley 1953). However, empirical results consistently show that actual behavior is inconsistent with these assumptions (Henrich 2000). Another assumption of game theory is that games are cooperative affairs (Nash 1950a, 1950b, Shapley 1953), though in many cases (and indeed in the game described in this thesis), they are not. In anonymous non-cooperative games, players do not know the identities of one another and are not allowed any communication. So players make their 'rational' decisions without consultation with other players, highlighting the individualistic nature of decisions (Gowdy and Seidl 2004, Roth 1997). However, Roth ( 1997) argues that decisions of players are also mutually interdependent, because the welfare outcome of (at least) one individual depends on 32

PAGE 46

the actions of another. This dependence sets the stage for strategy, and encourages (or forces) players to consider their actions in reference to outcomes for other players, as well as to anticipate their own outcomes at the mercy of other players. The acknowledgment of mutual interdependence15 distinguishes game theory from traditional economic theory. A final assumption, again stemming from canonical economics, is that culture is static or at least extremely slow to change (Gowdy and Seidl 2004). Economics can therefore handily dismiss culture and assume the universal homogeneity of human behavior in economic contexts. This model of universal economic man, termed "Homo oeconomicus," mirrors the biological model of man as innately self-concerned with fitness and reproductive success. Both traditional biology and economics limit individuals to fitnessor utility-maximizing endeavors, leaving little room for cooperation or altruism, unless as a means to gaining an advantage over some other individual (Gowdy and Seidl 2004). Ethnographic and empirical evidence from anthropology, biology, psychology, and behavioral economics shows that humans are not driven solely toward utility maximization. Despite its existence however, altruism remains outside standard welfare economic models. These varied but incomplete models fail to acknowledge the complicated nature of social interactions that involve competition and cooperation. In fact, game theory experiments often miss this as well, as laboratory experiments usually consist of isolated dyads acting in simulated conditions, with few available options and little or no communication between 15 Mutual interdependence is ignored in neoclassical economics because of assumptions about markets and market failure. Mutual interdependence also permits the condition of Pareto inefficiency, or the condition where no actor's welfare can be ameliorated without damaging the welfare of another actor (Debreu and Scarf 1963), a taboo condition in neoclassical welfare models which emphasize individualism and rational choice in producing ideal outcomes. 33

PAGE 47

actors. Nonetheless, game theory provides an important starting point from which to evaluate results from behavioral economic experiments. Debate about the applicability of predictions and generalizations of economic and game theory echoes the formalist-substantivist debate in economic anthropology (Burling 1962, Cook 1966, Dalton 1969, LeClair 1962, Po1anyi et al. 1957). In 1957, Polanyi et al. described the dichotomous meanings of 'economic.' The substantive definition of economic refers to the fact that individuals depend on both nature and peers for survival. That is, continuous interactions with his social and physical environment supply an individual with the means to satisfy his material wants and needs. The formal (and conventional) definition of economic refers to the logical and rational character of the means-end relationship described above, in which individuals choose between alternate means, some of which are better than others in the satisfaction of material wants. Scarcity is implicit in the formalist view. There is a finite supply of available resources for the satisfaction of wants; so while scarcity produces deficits in alternate means, and forces the making of rational choices in the formalistic view, scarcity need not be a factor in the substantive view. The formalist-substantive debate is an argument of whether the conventional, formalist view is helpful in the social sciences-specifically in the examination of non-Western, "primitive" cultures-as it is substantiated on the elemental assumptions of neoclassical economics (e.g., rational action, market behavior, and pricing). Substantivists (Dalton 1969, Polanyi et al. 1957) tend to emphasize induction and cultural relativism, and warn against asking questions based on our own (Western) economy. Furthermore, substantivists generally believe that the difference( s) in Western and nonWestern economies is one of kind while formalists believe the difference is one of degree. Formalists (Burling 1962, Cook 1966, LeClair 1962) emphasize deduction and the universality of economies, insisting that our own economy is a good reference point from which to compare other types of economies. 34

PAGE 48

Likewise, it could be argued that game theory, based on conventional economic theory, is not only unrealistic in industrialized nations (because it presumes that controlled, anonymous, laboratory interactions approximate reality), but it is especially inapplicable to studies in small-scale societies (because of the latter, in addition to presumptions about the universality of market economies). Like formalists however, proponents of game theory argue that theoretical groundings are a good starting point and greatly simplify analysis. The heuristic value of theory is not only made manifest when theory accurately predicts behavior, but when it illuminates anomalous behavior deserving of further study. 2.4.3 Games: The Prisoner's Dilemma, Dictator Game, and Ultimatum Game The prisoner's dilemma (PD), a model for political, business, and biological interactions (Bartholdi et al. 1986), is a static game-one in which all players act without information about other player action-based on the scenario16 of two suspects being arrested for the same crime without sufficient evidence to firmly convict either unless at least one confesses (Holt and Roth 2004, Romp 1997). If neither confesses, both will be convicted of a minor offense and sentenced to one month; if both confess, they will each be sentenced to six months; if only one confesses, the confessor will be released while the other will be sentenced to nine months, that is, six months for the crime and three months for obstruction of justice. In the experimental game version of the prisoner's dilemma, the two players are given the option to cooperate or defect with the monetary payoffs for each action loosely based on the punishment-reward scenario described above (Bartholdi et al. 1986, Holt and Roth 2004). Table 1 summarizes an example of such pay-offs. 16 The narrative ofthe Prisoner's Dilemma was devised by Albert Tucker, Nash's thesis advisor (Holt and Roth 2004 ). 35

PAGE 49

Table 1 Example of a Prisoner's Dilemma Payoff Scheme PI 8 ayer Cooperate Defect Player A Cooperate (80,80) (0,100) Defect (100, 0) (35,35) There are four possible outcomes: (cooperate, cooperate), (cooperate, defect), (defect, cooperate), or (defect, defect). Payoffs are in the format (Player A, Player B). The pure Nash equilibrium strategy is mutual defection; however, if both players cooperate, they are both better off than if they both defect (Romp 1997). If players are indeed selfish egoists, the preferred or most profitable outcome for each player is defection while the other player cooperates. As mentioned above in the section on reciprocal altruism, through several rounds of computer simulation, Axelrod and Hamilton ( 1981) showed that not 'always defect' (ALLD), but tit-for-tat (TFT) is the prevailing and most successful strategy in iterative rounds of the prisoner's dilemma. Thus the computer uses a strategy derived from past experience of opponent cooperation or defection (Bartholdi et at. 1986) instead of selfishly always defecting. Bartholdi et at. ( 1986) extend the findings of Axelrod and Hamilton ( 1981) to argue that not just individuals, but corporations and even nations conduct behavior using a TFT strategy. Once established, TFT is extremely robust and resistant to intrusion by mutant strategies. Leaving computer simulations behind, the following examples come from economic games performed among actual individuals. Empirical results from iterative rounds of prisoner's dilemma games consistently demonstrate that with time, individuals move from a high level of 36

PAGE 50

cooperation toward more frequent defection. For example, players usually cooperate 40-50% of the time in initial rounds, but cooperation drops to around 20% with repeated play (Romp 1997). Romp (1997) argues that this demonstrates increased game experience influences participants to choose cooperation less and less, thus tending towards the Nash equilibrium (ALLD). As Axelrod argues (1984), a more accurate statement might be that players adopt a TFT reciprocal strategy. They choose cooperation in the first round and then base subsequent decisions on experience. However, one-fifth of the participants still chose the cooperative strategy after several rounds, so it seems that the Nash prediction only receives limited support (Romp 1997), as does the assumption that individuals are rational (or will play rationally) and that they know their opponents are rational as well. Potential explanations for such altruistic behavior assert that players get additional utility from cooperation in one of three forms (Andreoni and Miller 1993). In (I) pure altruism the player is concerned not only for himself, but for the welfare of the other players, so that additional utility is received due to the greater gain for the other player; because of (2) duty the player cooperates from moral obligation, so that additional utility is received from cooperation; in (3) reciprocal altruism the individual receives extra utility when both players cooperate; this form is sometimes termed 'warm glow,' as mutual cooperation is supposed to convey pleasure for both players17. In other words, payoffs do not just consist of the money doled out by the experimenter, but include intangibles. This rationalization suggests that seemingly irrational behavior may in fact be rational and utility-maximizing, when utility is considered in non-monetary payoffs. 17 Though previously discounted, this idea may receive additional support in the near future from neurological studies of the basis of altruistic punishment that assert that not strategy, but 'good feelings' drive the punishment of norm-violators; in other word, altruistic behavior, including cooperation and the punishment of defectors is psychologically rewarding, and may thus be evolutionarily-based (Fehr and Rockenbach 2004, Quervain et al. 2004). 37

PAGE 51

Another simple and frequently performed game is the dictator game (DG). In the DG, Player A, the dictator, is given a sum of money that he may divide between himself and Player B. Player B is inert and merely receives whatever, if any, payoff Player A gives In non-anonymous games, the average amount shared is 50%; when player identity is kept anonymous, the amount shared drops to 36%, a percentage much higher than the expected offer of 0% should players act according to selfish motives. Even more astounding are results from the ultimatum game, a variation of the DG that adds the element of punishment by Player B. The ultimatum game (UG) consists of an exchange between two players, with a potential gain for both if the players can agree on how to divide up the sum (Camerer 2003, Fehr and Gachter 1998, Gowdy and Seidl 2004, Henrich 2000, Nowak et al. 2000). Player A, the ultimatum-giver, proposes a take-it-or-leave-it offer of how to divide up the endowment with Player B. If Player B agrees, Player B receives the proposed amount while Player A takes the remainder; if Player B does not accept the proposal, neither player enjoys any gain19. There is no negotiation. Typically, like the PD and DG, payoffs are given in real money and the players are anonymous to everyone except the experimenter20 Game theory (via economic theory) predicts that players (or more generally humans, H. oeconomicus) are selfish and rational so that in the one-shot ultimatum game, Player A will offer as little as possible and Player B will accept any non-zero offer (Camerer 2003, Henrich 2000, 18 Because only one player has the opportunity to make a decision and act, some argue that the DG is not really a 'game.' 19 For example, Player A is given $10, and must offer some amount, x, to Player B. If Player 8 agrees, he accepts x, while Player A keeps $1 0-x. If Player B rejects the offer, neither player receives any money. 20 Some experimental economists (Hoffman eta!. 1994) arrange the game so that even the experimenter is anonymous, to control for experimenter-influence of player behavior. Hoffman eta!. (1994) highlight the potential for costly signaling by players that may promote altruistic behavior (i.e., the player wants to appear generous to the watchful experimenter, and thus proposes a higher offer, despite his true sentiment). 38

PAGE 52

Gintis et al. 2003, Gowdy and Seidl 2004, Nowak et al. 2000, Sigmund et al. 2002) because 'something is better than nothing However, Tracer (2003) predicts that Player B will not accept anything less than a 50-50 split because an unequal split might produce an imbalance in relative fitness, under the simplifying assumption that each unit of payoff is convertible to a unit of fitness. UG experimentation with university students in numerous culturally and geographically diverse developed countries seems to disprove the above economic theoretical assertion that players are inherently selfish. The modal Player A offer is consistently 50%, and the mean offer between 40% or 50%; offers of less than 20% are refused about half of the time regardless of sex, age, and degree of anonymity (Camerer 2003, Roth et al. 1991, Henrich 2000, Gowdy and Seidl2004, Nowak et al. 2000, Roth et al. 1991, Sigmund et al. 2002). Arguably, the same results occur when the stakes are higher21, that is, in low-income developed countries, where the stake is equivalent to several weeks work of wages (Cameron 1995, Fehr and Gachter 1998, Fehr et al. 2002, Gowdy and Seidl 2004, Henrich 2000). It seems, then, that players are motivated by some sense of fairness independent of their own payoff; or that they are willing to (altruistically) decline offers of 'free money' in order to punish other players for their unfairness. Fehr and Gachter ( 1998) assert then, that not H. oeconomicus, but Homo reciprocans, is the more accurate taxonomic designation of players (Sigmund et al. 2002). Rejections of non-fair splits, or non-50-50 offers, may be an example of strong reciprocity (altruistic punishment), where Player B reciprocates the unfair behavior doled out by Player A even though retribution comes at a cost to himself. The important question becomes the origin of this principle of fairness. As mentioned above, Tracer (2003) argues that 'fairness' (equal split) is 21 This does seem to work with relatively higher stakes in under-developed countries. However, Telser (1995) shows that when the stakes are very high, rejection is very low or non-existent even if offers are as low as one percent. For example, if the original endowment is $1,000 and the offer is $100, Player B is very unlikely to reject. Thus, following the law of demand, the split approaches extreme inequality with increasing endowments. 39

PAGE 53

congruent with the other-regarding, relative fitness-maximizing predictions of evolutionary theory. Recent behavioral economic experiments by eleven anthropologists in Africa, the Amazon, Papua New Guinea, Mongolia, and Indonesia demonstrate that in these small-scale societies (foraging, horticulturalist, nomadic herding, and small-scale agricultural societies), results are very different from those in developed countries where games were typically played with university students (Bolton et al. 1998). In an anonymous UG played among the Machiguenga of the Amazon, Player A offered very little (mean offer: 26%; mode offer: 15%) while Player B consistently accepted almost every offer, even those less than 20% (Henrich 2000). These offers and rejection rates are much lower than those found among university students. Conversely, among the Au of Papua New Guinea, Player A sometimes offered more than 50%, some citing fear of village turmoil should any (too-low) offer be refused (Henrich et al. 2001, Tracer 2003, 2004 ). In the above examples, the experimenters concluded that culture was the primary factor determining behavior, including expectations of proper offers and beliefs about fairness. In the Papua New Guinea case, 'over-sharing' was explained by cultural rules about generosity, exchange, and reciprocity, such as those that prescribe that hunted game cannot be consumed by the hunter and must be shared. However, responders in Papua New Guinea rejected hyper-fair offers as well, allegedly due to cultural rules about competitive gift-giving. These rules dictate that taking a large sum incurs even greater future obligation. Indeed, the Au tended to reject both excessively generous and low offers equally. Henrich (2000) asserts that the Machiguenga felt neither an obligation to divide the endowment equally, an expectation to receive an equal share, nor an innate desire to punish an unequal division. In other words, they felt the modal offer of 15% was 'fair.' In sum, results from these experiments in small-scale societies show that 1) the canonical economic model of self-interest is not supported; 2) individual-level demographic and economic variables do not explain behavior within or between 40

PAGE 54

groups; 3) there is great behavioral variability cross-culturally, but behavior is explained in part by degree of market integration22 (hence the difference in results between university students in developed nations and in participants from small-scale societies); 4) local economic patterns are generally congruent with degree of cooperation (i.e., economies of scale in production) and/or punishment. It seems then, that different cultures produce different standards of fairness, often consistent with degree of anonymous market exchange (Camerer 2003, Henrich et al. 2001, Tracer 2004). Also, with an increase in market integration and Westernization, cultures seem to have sharing norms that produce more equal splits. In other words, in cultures with the most market integration, offers seem to be the least selfish (Camerer 2003). But is it that market experience creates regulating norms about fairness and equal division, or that the proclivity toward even distribution produces a favorable market environment? Offers and rejections in cross-cultural ultimatum games may be said, then, to be a sort of language portraying cultural nuances. For both developed and non developed countries, UG results suggest that many players play by cultural rules according to what is locally 'fair' (Clark and Sefton 2001, Eckel and Grossman 1998, Schroeder et al. 2003). These may include reaction to unfairness I) in outcome (Bolton and Ockenfels 2000, Fehr and Schmidt 1999), 2) in intent (Rabin 1993), or 3) in both outcome and intent (Charness and Rabin 2002, Falk and Fischbacher 1999, Falk et al. 2003). The first is based on the assumption of a utility function, and refers to equitable outcomes-in game theory, payoffs-as based on some measure of how close an actual outcome or offer approximates a fair reference pay-off. Theories of fairness in outcome may help explain the one-shot prisoner's dilemma, public goods, gift-exchange, and ultimatum game rejections (Diekmann 2004). The second and highly rational theory assumes that stakes matter, and refers to an evaluative 22 Market integration refers to an index including existence of a national language, existence of a labor market and farming for cash (Camerer 2003). 41

PAGE 55

response in kind for intentional23 behavior; that is, 'eye for an eye,' whether the 'eye' is benevolent (fair) or vengeful (unfair). This theory may again explain ultimatum game rejections. The third argues for the importance of fairness in both intent and outcome. Still others (Charness and Rabin 2002) claim that punishment may be explained by competitiveness of individuals who prefer to have their payoff be as high as possible relative to other payoffs. This theory sounds strikingly similar to that of other-regarding relative utility maximization (Tracer 2003). It has been argued that humans universally value the emotional and moralistic idea of fairness (Sigmund et al. 2002). Sigmund et al. (2002) propose that humans evolved the emotions at the heart of this situation from life in small groups, where behavior that not only benefited individuals but that benefited the group in the long run was favorable. The authors suggest that fairness is a sort of self-conscious behavior, because others (in our group) are watching and remembering what we do, and thus may predict what we are likely to do in the future. In other words, we all have and must maintain a reputation. One-shot encounters were infrequent throughout human evolution, so that individuals tend to respond angrily to defection in an effort to nurture both self-esteem and reputation, as an individual that will not tolerate being cheated. While it could be argued that humans indeed have something called 'good character,' and that they truly enjoy helping and sharing with others, surely there must be a biological component to fair behavior (Sigmund et al. 2002). Sigmund et al. (2002) argue that altruism, and other social emotions like friendship, guilt, resentment, and shame, help us negotiate (ultimately biological) success in complex social networks. Results from dictator and ultimatum games as well as more complicated game experiments (public goods games, gift-exchange) consistently show that 23 Trivers ( 1971) also pointed out the problem of intent when defining altruism, that is, whether to identity altruism by motive or by behavior regardless of motive. 42

PAGE 56

humans are not primarily selfish. Theoretical economic and evolutionary views of behavior, then, are poor predictors in lived realities where cultures and social norms play a much bigger role than previously allowed in the traditional view of H. oeconomicus, the rational and self-interested actor (Fehr and Gachter 2002, Gowdy and Seidl 2004, Henrich et al. 2001 ). This suggests there is a need for a new behavioral economics (Gowdy and Seidl2004). In order to further the debate about the roles of altruism, fairness, and justice in game behavior, this thesis reports the results of a modified dictator game. As in the dictator game, a proposer decides how he and an anonymous second party will divide an endowment. Then an anonymous third-party is given an endowment equal to half of that divided between the other two players; he is also give the opportunity to usurp the power of the dictator and change the proposal. He may sanction the proposer, compensate the recipient, or both punish and compensate, but all at a cost of a percentage of his endowment; or, he may take his full endowment and leave the payoffs as they stand. From a rational and selfish perspective, third-party punishment or compensation is altruistic and unjustified: it reduces the utility of the third-party. 2.4.4 Weaknesses of Game Theory Ideally, game theory has the potential to provide a holistic and realistic window into economic behavior with the potential to elucidate understanding of the evolution ofthat behavior (Gowdy and Seidl2004). However, game theory has been criticized for its simplicity, abstraction, lack of real-life haggling and bargaining, lack of face-toface contact, lack of a feed-back loop, rigid structure that disallows mutation in the form of player ideas, limited time for making evaluative decisions, lack of realistic context, and typical limit to one-shot non-repeated encounters (Camerer 2003, Gowdy and Seidl2004, Romp 1997). For example, the prisoner's dilemma has been criticized for predisposing players toward selfish action because players may not communicate, cheating is rewarded, and the punishment of free 43

PAGE 57

riders is not possible. Such lack of autonomy and communication denies players of a primary and perhaps unique human attribute (de W aal 1996). Nonetheless, game theory remains attractive in the social sciences for the potential of quantification of the intractable subject of human behavior (Sigmund et al. 2002). In summary, simple experimental economic games like the ultimatum game test game-theoretical principles, and are useful to understand what people think about the allocation of resources to both themselves and their peers (Camerer 2003, Fehr and Gachter 1998, Sigmund et al. 2002). Game theoretical experiments consistently show that when given the opportunity, players will respond in kind (Rabin 1993), rewarding those who are generous and punishing those who are not (Fehr and Gachter 1998). However, they also tend to play games according to socially and culturally informed rules of fairness. In addition, since most experiments are anonymous, reciprocity and punishment seem to apply even when people do not know with whom they are dealing. 2.5 Conclusion Since the publishing of Adam Smith's24 Wealth of Nations, standard economic theory has assumed that all humans are in the business of maximizing their self interests (Balasko 1988). In the jargon of economics, when consumers are faced with the choice between two commodities, they are sentient and rational enough to compare the two and choose the preferred commodity based on preference (Balasko 24 .. but man has almost constant occasion for the help of this brethren, and it is in vain for him to expect it from their benevolence only. He will be more likely to prevail, if he can interest their self love in his favour, and show them that it is for their own advantage to do for him what he requires of them .... Whoever offers to another a bargain of any kind proposes to do this. Give me that which I want, and you shall have this which you want, is the meaning of every such offer; and it is in this manner that we obtain from one another the far greater part of those good offices which we stand in need of. It is not from the benevolence of the butcher, the brewer, or the baker, that we expect our dinner, hut from their regard to their own interest. We address ourselves not their humanity but to their self-love, and never talk to them of our own necessities but of their advantages" (Smith 1868:6-7) (emphasis added). 44

PAGE 58

1988). And we might infer that this preferential choice should be based on the drive to maximize self-interest. This idea mirrors the assumption of canonical evolutionary theory that all organisms are self-regarding, with the aim of maximizing individual fitness and subsequently reproductive success. According to the evolutionary theories of kin selection and reciprocity, cooperative behavior is favored by selection only when cooperators are more likely to interact with other cooperators; kin relationships seem to provide the most likely source of this (non-random) interaction (Boyd and Richerson 2005). Human eusociality may thus be explained by the combination of this reciprocity and the (perhaps unique) human ability to use our big brains to recognize a large number of individuals and to remember a large number of (historical) social interactions. These, coupled with punishment as a disincentive, may be the keys to understanding human cooperation. Hamilton's idea of kin selection is the most convincing model for the inception of altruism. Like primate groups, the first groups of humans were primarily made up of kin, with some mixing of lines to avoid inbreeding. Cooperation first arose, as seen among non-human primates, as a reproductive strategy for group living, to avoid predators and to permit the best utilization of resources. As intelligence increased and social organization grew more complicated, cooperation too grew increasingly more complicated to ensure social bonds. Cooperative ties and reciprocal acts began to extend to non-kin, as encouraged by previous positive experiences with relatives. As interactions increasingly took place between distantly and non-related individuals, natural selection favored the ability to both remember a large number of past interactions and to discern honesty. Negative reciprocity, also termed altruistic punishment, naturally flowed from positive reciprocity, as a deterrent for defectors. The theories ofboth direct and indirect reciprocity and of altruistic punishment, explicate what happens once cooperation comes into vogue, better than they explain how cooperation could arise and evolve. From Hamilton's 45

PAGE 59

theory, the extrapolation of the benefits of altruistic cooperation can be synergistically extended. Reciprocity and punishment are sustaining forces after cooperation is already in place, especially in large groups with highly specialized division of labor. Thus kin relations first produced cooperation; punishment sustains it among non-related individuals. Though primatologists, biologists, and anthropologists have long dismissed it, explorations of altruism often return to the controversial ideas of group and multi level selection because kin selection and social reciprocity fail to fully explain altruistic behavior (Bradley 1999, Darlington 1972, 1978, McAndrew 2002, Smith 1976, Strier 2000). There is evidence for group selection among virus strains and foraging ant queens (Bradley 1999). By nature, altruism occurs in groups, even if only groups of two (Darlington 1978). If selection occurs at the level of the group, then groups with altruists should have higher fitness than those without them (Strier 2000). But in nonhuman groups, it seems that the animals who warn others of predators increase their chances of detection and thus decrease their fitness; in addition, selfish individuals would pass on more genes by free-riding on such warnings. However, if a distinction is made between individual selection, which acts on differential fitness of individuals within groups, and group selection, which is the same principle acting on differential fitness of individuals between groups, one altruist surrounded by a horde of selfish individuals will have lower individual fitness; but a group of cooperative and altruistic individuals may have higher fitness than individuals in other groups composed of mostly selfish members. If a multi level hierarchy of selection pressures acts on individuals, on populations, and on species, etcetera, and differences are heritable at each unit of selection, natural selection could conceivably operate at multiple levels in a sort of co-evolution (Darlington 1978, Strier 2000). Strier (2000) cites the example of a selfish group that overexploits its niche, thereby reducing the average fitness of all group members. In juxtaposition, cooperative members of a largely altruistic group may have lower 46

PAGE 60

individual fitness than selfish members of other groups, but over time will persist longer than the selfish group. Though selfish members could infiltrate an altruistic group, if punishment becomes the norm, their selfishness can be stifled. Darlington ( 1978) agrees to a degree, but argues that though altruism is a group issue, it evolves by individual selection, is opposed by competition, costs, and inefficiency, but is supplemented by group selection in a multi-level process. Though these issues are very difficult to study among nonhuman animals, game theory provides an important foundation from which to build explorations of altruistic cooperation. Is it an evolutionary accident or even maladaptation (Fehr and Henrich 2003) that humans cooperate, especially with strangers with whom they meet only once (Gintis and Bowles 2003)? Is this ability the result of the uniquely human qualities of language or cognitive ability, or the cultural tendency to have prosocial norms and institutions that guide our social conduct toward fairness, including the tendency to punish violators? Humans often live and work in groups of non-kin. Is cooperation a necessary pre-condition, or by-product of this arrangement? A substantial body of evidence supports the idea of kin selection among cooperating animals; a less robust body of evidence supports reciprocity as an explanation of animal, specifically primate, interactions. These theories do well to explain human interaction with kin and individuals with whom humans repeatedly interact, but costly signaling and strong reciprocity (altruistic punishment) carry evolutionary theories further, yet not far enough to explain one-shot altruism. The interaction of both genes and culture are necessary to understand altruistic cooperation in an evolutionary framework. Perhaps the idea of multi-level selection on variability in both genes and culture provides the key to understanding altruism (Boyd and Richerson 2005). Though existing theories do an adequate job of explaining cooperation among kin and for non-kin who interact repeatedly and frequently, explanations of one-shot altruism remain unclear so that further empirical testing with the aim of refining theory is required. 47

PAGE 61

3. Study Population 3.1 The Setting 3.1.1 Papua New Guinea Papua New Guinea (PNG) is located just north of Australia along the "Ring of Fire" and shares the second largest island in the world with the Indonesian region of West Papua, formerly called Irian Jaya (See Map A. I). The 820 km land boundary with Indonesia not only divides the island of New Guinea in half vertically, but separates the two nations culturally. PNG is most often grouped into the South Pacific islands known as Melanesia, along with its nearest neighbor to the east, the Solomon Islands, and Fiji, New Caledonia, Vanuatu, Maluku, and the Torres Strait Islands. West Papua became a part of Indonesia in 1962 after the Dutch lost administrative control (Tracer 1991 ). From 1885 until 1902, control of the eastern half of New Guinea was maintained by Germany in the northern Territory of New Guinea and the United Kingdom in the southern Territory of Papua. In 1902, Australia took control of the southern territory on behalf of the United Kingdom (Banks 1998); subsequently, during World War I, Australia gained administrative control of the northern territory from Germany as well (Tracer 1991 ). Australia maintained control until 1975 when the nation state of Papua New Guinea achieved independence. PNG's geo-political situation changed once again in 1997 with the secession of Bougainville, the largest of the Solomon Islands, following a bloody nine-year revolt (Banks 1998). 48

PAGE 62

Geographically, PNG boasts 5,152 km of coastline and a total area of 462,840 sq km25, an area slightly larger than the state of California (CIA 2005). The rugged, tropical terrain is mostly mountainous with coastal lowlands and rolling foothills, rendering only 0.46% of land arable for agriculture. The highest point is Mt. Wilhelm at 4,509m. Though PNG possesses considerable natural resources including gold, copper, silver, natural gas, timber, oil, and fisheries, the rugged terrain and lack of infrastructure-the total length of roads in PNG is 19,600 km, with only 686 km paved-prevent major exploitation of these resources (to the delight of travelers and anthropologists alike). Still, 72% of export earnings come from mineral deposits including oil, copper, and gold. As a result, pollution from mining and deforestation are the primary environmental threats. Finally, 20% of the national budget is derived from the $240 million in aid received annually from Australia. In 2000, the PNG census reported a population of 5,190,786 people (PNG NSO 2005), with current estimates (July 2005) of 5,545,268 citizens. The population has a median age of21 years with an average life expectancy of65 years26 The current total fertility rate is 3.96, but this figure varies greatly by region. Eighty-five percent of the population survives on a subsistence lifestyle of gardening, animal husbandry, and limited cash cropping (Banks 1998) including coffee, cocoa, and increasingly vanilla. The social organization is what Henrich (2000) has called a family-based society, meaning that extended families, or clans, predominantly produce for themselves and do not rely on institutions for their welfare. In such societies, anonymous transactions are almost unknown, and a high degree of reciprocity pervades daily life. Though English is the national language of PNG, only 1-2% of the population is fluent. The linguafranca ofPNG is Melanesian pidgin (Tokpisin), but an 25 Including 452,860 sq km land area and 9,980 sq km water area. 26 Male and female life expectancies are 62.76 and 67.21 years, respectively (PNG NSO 2005). 49

PAGE 63

estimated 700-800 indigenous languages are spoken in PNG, many of them unrelated (Banks 1998, Tracer 1991 ). The lack of infrastructure and relative isolation mentioned above not only currently restricts development, but helped create and continues to maintain the cultural differences and linguistic diversity found among groups even when the absolute distance between them is minimal. Papua New Guinea is a constitutional monarchy with parliamentary democracy (Banks 1998). The legal system is based on English common law. This unitary system, a vestige of Papua New Guinea's colonial past, provides that while PNG's 19 provinces and national capital district have local legislative power, provincial legislation may be vetoed by the creation of National Acts (Banks 1998). The official criminal justice system of PNG is also a remnant of colonization. Western assumptions about guilt, innocence, retribution, and individual responsibility pervade the imposed criminal justice system, making it often at odds with traditional ideas about principles of justice and the methods used to keep it. The disconnect in ideas about justice, the language used to describe it, and the protracted action of the courts in trying cases leave the judicial system largely incomprehensible to many indigenous peoples who generally believe in swift reprisal to address violent crime, and largely emphasize victim compensation over other punitive measures (Banks 1998). 3.1.2 Anguganak, Sandaun Province This study was conducted among the Au of Anguganak, Sandaun27 Province, in northwest Papua New Guinea. Anguganak comprises a series of hamlets about 50 kilometers inland from the northern coast and 95 kilometers east of the Irian Jayan border (Tracer 1991) (Map A.l). The Au occupy the southern foothills of the Torricelli Mountain range, living at altitudes of 150 to 850m. This mountain range 27 Formerly West Sepik Province. 50

PAGE 64

isolates the Au from the Pacific Ocean. The Au populate lowland tropical rain forest, typical of an equatorial climate where temperatures vary little annually. Annual rainfall is very high, exceeding 2.5m. Though the periods from October to March and from April to September are classified as the wet and dry season respectively, the dry season is not typified by drought but a reduction in overall rainfall with accompanied drops in river and spring flow. Malaria is endemic to the area as the climate encourages the proliferation of anopheline mosquitoes; indeed, malaria is the preeminent cause of both child and adult mortality. Other health concerns include dengue fever, tuberculosis, tapeworm, tropical ulcers, and scabies. 3.2 The People and Culture 3.2.1 The Au 'Au' not only refers to the approximately 10,000 individuals who occupy about 50 villages in the East Au and West Au census divisions, but is also the name of the predominanrx language spoken in the area (Tracer 1991 ). Villages, generally constructed atop ridge lines, range from less than 100 to more than 500 people. The spatial layout of the villages, constricted by the width of the ridge line, usually consists of several ham lets strung together by mudstone paths. Most likely the result of missionary influence, the current housing layout and therefore sleeping arrangement is very different from that described by Lewis (1980) and Tracer (1991). Men's houses that used to provide sleeping quarters for all men and most boys over the age of 10 have disappeared. Nuclear families now reside together, some in ground-level, windowless, thatched-roof houses; but many live in more modem 28 Other languages are spoken within the East and West Au census boundaries, including Gnau (Lewis 1980), Elkei, Ghal, and Yil, which together make up an ancient and unique phylum of languages ostensibly distinct from other languages outside of the Sepik region, and arguably may provide a link to the original languages brought to PNG by incipient immigrants arriving from the Malay area (Tracer 1991 ). 51

PAGE 65

houses built atop stilts, some with screen-covered windows, multiple rooms, and increasingly, corrugated iron roofs (Photographs A. I, A.2, and A.3). Tracer (1991, 2003) characterizes the Au as forager-horticulturalists, surviving primarily on jelly made from the pith of the sago palm and leafy greens such as the 'jointfir spinach' (Gnetum gnemon) (Photographs A.4, A.5, A.6, and A.7). The Au supplement their diet with other vegetables grown in their slash-andbum gardens, including taro, sweet potatoes, bananas, pandanus, amaranth, and papaya; with gathered foods like wild mushrooms, breadfruit, nuts, grubs, and insects (mainly eaten by children); with animals like snakes, lizards, birds, and bird-eggs happened upon during daily activities in the bush; more rarely with hunted game including bandicoot, wild pig, and flying fox; and still more rarely (usually on prestigious or ceremonial occasions) with domesticated animals, such as pigs and chickens. Unlike others areas of PNG like the highlands popularized by Rappaport (1968), pigs are not abundant in Anguganak. In fact, only two domesticated pigs were observed during the entire field season. Fish are rarely eaten unless they are of the store-bought tinned variety, as the nearest rivers have few large fish29 Finally, if they can afford ie0 the Au supplement their diet with store-bought rice and instant noodles, purchased at the trade-store on the mission station (hereafter called "the Station") set up alongside the Anguganak airstrip (Photograph A.8). The two small (one single-engine and one twin propeller) planes that arrive at the airstrip twice a week provide the most reliable connection with the nearest port town ofWewak, as the arduous road between the two is unpaved, often impassable, and sometimes dangerous due to both road conditions and bandits. The Station is also the location of 29 However, plans are being made in at least one Au village to import tilapia and stock two dug-out fish ponds that, currently empty, are better classified as mosquito farms. 30 Those who buy food are usually wage-earners employed by either the government, for example, as nurses or teachers, or by missionaries. Wage-earners are decidedly in the minority, as for example, only 11.8% of our sample works for wages. 52

PAGE 66

the public school that all children have the opportunity to attend if their families can afford the enrollment fee (Photograph A.9). Though recent data could not be found, old data on health indicators in Anguganak are interesting and probably comparable to current numbers31 The infant mortality rate reported by Tracer ( 1991) more than a decade ago was 104/1000 live births, a number unchanged from previous studies up to two decades earlier. The mean age of marriage for girls was about 21 years, with first birth occurring about 2 years later. The total fertility rate, defined as the mean number of live-births ever experienced by post-reproductive aged women (over the age of 45), was 6.1. This number is comparable to other 'natural fertility populations,' or populations that do not have the intent or the means to control parity (Tracer 1991 ). 3.2.2 Reciprocal Exchange Like most Melanesian societies, a complex and rigid system of reciprocal exchange pervades the social, economic, and political sectors of the lives of the Au, for example in social grooming-delousing (Photograph A.1 0), food taboos-a hunter may not consume his kill, but must distribute it among kin-and in marriage practices (Banks 1998, Lewis 1975, Sillitoe 1998, Tracer 2003, see Zimmer Tamakoshi 1997 for a detailed account of exchange rules, especially with regard to land tenure and power relationships). Not only does a man's family pay a brideprice for his wife, other payments are made to her family at the birth of the first child, the child's first consumption of meat, puberty, et cetera. In exchange, the woman's brother (the child's maternal uncle) nurtures a special relationship with his niece or 31 Due to the lack of development and therefore lack of medical advancement experienced by the village, health indicators probably have not changed drastically over the past two decades. In fact, Anguganak may be less-developed now than it was in the past because oflack of full-time missionaries and clinic personnel from other countries, and fewer missionary flights to the area with supplies. The Station was previously home to a post office and a bank, and missionaries installed amenities such as phones run on electricity from generators. All of these are now gone, save the generator at the medical clinic. Radio is the only means of communication with Wewak. 53

PAGE 67

nephew32 providing them with protein, performing ceremonial rites, and so on. Failure to comply with these social norms may result in ostracization or violence. In addition to formal reciprocal mores, the Au consider it a right to request everyday items from each other, most frequently betel nut and food, but also more valuable items like clothing, string bags, tools, and even money (Tracer 2003). If X requests an item, Y must comply. IfY refuses, she risks being shunned, physically abused, or at least being ignored should she request an item of someone in the future. However, should X abuse her rights to request items and do so too frequently, she may also be shunned or peppered with requests for items. In addition to generosity with solicited items, the giving of unsolicited gifts, usually of hunted game or other food, also serves to strengthen social ties (Sahlins 1972, Tracer 2003). Taking a gift of either kind binds an individual to return the favor at some time in the future, upon request or otherwise. While the generosity norm keeps wealth and goods evenly distributed, it also encourages discreetness with goods and hunted game. Moreover, because the Au recognize their future obligation when receiving a gift, they sometimes refuse offers due to unwillingness or inability to pay it back (Tracer 2003). Finally, compensatory gift-giving after wrongdoing helps to adjust, maintain, restore, redefine, or in the case of inadequate compensation, break relationships (Banks 1998). 3.2.3 Law and Justice As a vestige of colonization, the relationship between formal law and custom has been (and still is) strained in PNG. The imposed western-based, judicial system 32 For example, during the field season, a seven-year old child standing under a coconut tree was struck on the head by a falling coconut and died. Members of the child's village and her mother's natal village went into mourning, and a series of compensatory exchanges ensued. Most interesting was the hefty monetary compensation paid by the child's father, his family, and other village members to the child's maternal uncle, in apology for not taking better care of his niece. The uncle reciprocated by hosting a feast for the mourners. 54

PAGE 68

is largely incongruous with internal belief systems and cultural notions. Formal codes do not take into account customary beliefs about dispute settlementespecially ideas about swift and violent reprisal for crime, and about victim compensation-instead imposing Western assumptions about universality, individual responsibility, guilt, innocence, and the necessity for a protracted trial system. The enactment ofthe Criminal Law (Compensation) Act in 1991 was an attempt to reconcile this incongruity, and to produce a system of law that integrates both the Western-based criminal justice system and traditional law (Banks 1998). In accordance with customary practices of victim compensation, the act empowers the national and district courts to order the guilty to pay compensation in addition to, or in lieu of, other punitive measures. Still, the geographical isolation of cultural groups with greatly differing customs, beliefs, and lifestyles, makes a blanket national law and court system nearly impossible to apply. The compensation act has been argued to convert violence, injury, or loss into currency so that wealth items may be accepted as equal return for the wrong (Banks 1998). Though the compensation act is not determinate but procedural, the courts usually treat the payment of compensation as a mitigating factor in punitive sentencing. Compensations are determined by the courts according to what each party (and his family or clan) is willing to give and accept. As qualitative results will show, compensation in part or in whole makes up what individuals deem as the appropriate punitive measure for both petty and violent crime. Compensation comprises two parts, the first being reparation of the crime and the second a reevaluation of the wronged relationship in a public manner. This echoes the principle of restorative justice, explored in the next chapter, which not only requires retribution by the offender but also engages the three actors involved in a crime-the offender, victim, and the community-in a joint attempt to amend the wrong. Ideally, the restorative process works to give a voice to actor feelings and 55

PAGE 69

issues. This includes public acknowledgment of shame of the offender and forgiveness by the victim (Cohen 2001). 56

PAGE 70

4. Justice 4.1 Types of Justice 4.1.1 Retributive Justice One modem form of state-controlled reciprocal altruism is the retributive criminal justice system. Retribution is a manifestation of the old adage 'an eye for an eye,' but the fact that retribution can refer to either positive or negative returns is often overlooked in favor of the punitive sense of the word. In the retributive paradigm of justice, the focus is primarily on the criminal act (Bennett 2002, Daly 2000), secondarily on how that crime represents a violation of the state, and lastly on the actual harm done to the victim or community (Zehr 1985). Presumably on behalf of victims but essentially on behalf of wrongs committed to itself, the state takes action against the criminal and should he be found guilty, determines an appropriate punishment in order to replace one social injury with another, allegedly to deter future crime as well as shore up public well-being (Hampton 1984). Punishment is usually imprisonment (Johnstone 2003) whereby alienation is both a sacrifice and a symbol of the community's moral disapproval. In other words, like Foucault's asylum, the prison is at once a tool of moral uniformity and social ostracization (Rabinow 1984). Ideally, the imprisoned wrong-doer will proceed through an emotional journey from guilt to shame (a form of self-imposed punishment) (Bennett 2002). Imprisonment is based on the reform movement championed by the Quakers in the early 1800's, wherein criminal behavior was thought to be the result of the 57

PAGE 71

criminal's corrupt social environment (Griset 1991 ). To be effective, imprisonment should rehabilitate, deter, and incapacitate. That is, it aims to produce Foucaultian 'docile bodies,' stripped of their destructive 'power' by the disciplinarian state and coerced into the form of a 'normalized' citizen (Foucault 1977). Fundamental to the effectiveness of this retributive system is stigma that ideally should shame the perpetrator into reform and discourage others from emulating his criminal behavior. However, in most societies, stigma lingers long after the actual imprisonment, effectively ostracizing the criminal and preventing reintegration through post imprisonment limits. These might include the inability to obtain a driver's license, difficulty in getting a job, denial of rights including the right to vote, etcetera (Johnstone 2003). While effective and re-integrative shaming could prevent recidivism as well as keep others from imitating criminal acts, stigmatization more often exiles, humiliates, and may even perpetuate crime (Braithwaite 1996, Cohen 2001, Foucault 1977). Zehr (1985) argues that throughout history, there has been a dialectic between two forms of justice: that of the state and that of the community. Whereas state justice was legal, formal, rational, rigid, and punitive, community justice was flexible, context-dependent, often negotiated, and frequently restitution-oriented. Modem (and usually Western) interpretations of justice have taken the state model wholesale, simultaneously enhancing the central power of the state as the primary actor and using the prison as its punishment of choice; this is likely no coincidence. The imbalanced focus on pure punitive measures leaves little or no room for victim compensation. Punishment and restoration are inherently different, as punishment is a determinate means toward reform while restoration is a flexible process with a potential outcome; punishment is probably not the most effective means to the end of restoration (Walgrave 2004). Retribution is born of moral and ethical sentiments (Hampton 1984 ), not unlike the principle of fairness. Under conditions of punitive 58

PAGE 72

retribution, the offender suffers in what is deemed an equal amount to the victim; he pays the victim back in suffering so that the amount of suffering is doubled and spread 'equally' between the two (Walgrave 2004). In the restorative justice paradigm, the offender actively pays back in reparations, constructively taking suffering away instead of adding to it (or having it added to on his behalf). 4.1 .2 Restorative Justice Restorative justice is a progressive alternative to retributive justice, born of dissatisfaction with exclusionary means of crime control, particularly crowded and seemingly ineffective prisons (Cohen 2001). In lieu of the punishment-focus of retributive justice, restorative justice focuses on perpetrator responsibility and damage reparation after a crime (Johnstone 2003). Control of the reparation process does not lie solely in the hands of a formal, state-run judicial system, but is also appropriated in the hands of community members including the victim and perpetrator of the crime. Ideally, all stakeholders--{)r everyone affected by the crime including community members----<:ollectively decide how best to resolve the harm done by the crime to victim, offender, and community, and how to prevent recidivism. State officials and justice agencies act as facilitators of the process. In this way the common layperson may "do" justice through a variety of methods. One of these is the mediation session currently used in Australia and New Zealand for youth crimes (Daly 2000). The victim and offender communicate directly in a public setting, and both participate in decision-making, as opposed to sitting inert and voiceless in a court room while professionals handle all aspects of the transgression and its reparation (Cohen 2001, Johnstone 2003). Main goals ofthe restorative process include I) healing: of the offender who must seek forgiveness (from victim, community, and self); ofthe victim; and of the community at large, whose members may feel unsafe or betrayed as a result of the crime; however, harm done to the community is of secondary importance to the direct victim; 2) encouraging the 59

PAGE 73

perpetrator to take responsibility for his actions; 3) emotional journey of the perpetrator from shame to gui It, to regret, and to empathy for the victim and community, therefore bolstering his tie to the community so that he may be fully reintegrated; 4) identification ofthe social psychological issues that all actors have to confront in order to reclaim the dignity of the wronged and shamed parties so that all can effectively reintegrate. In this process, remorse of the perpetrator is a key to restoration, as is forgiveness (Cohen 2001, Zehr 1985). Restorative justice is processual, not a tangible or fixed outcome. There is a tendency to associate the restorative process strictly with restitution made to the victim (Barnett 1977). When found guilty, the perpetrator may have the option to offer some sort of compensation to the victim, though not necessarily a monetary remuneration as it is so difficult to attach a dollar sign to trauma (Johnstone 2003). Other than money, reparation may also include work for the victim or community (in some form meaningful to the victim) or course attendance (counseling, anger management, Alcoholics Anonymous, etc.). Of primary importance is apology, followed by tailored reparation developed from the needs, both material and emotional, of the parties involved (Marshall 1998). Retributive compensation, especially when voluntary, is both symbolic and therapeutic: symbolic as a sign that the perpetrator is accountable for his actions; therapeutic as a potential alleviator of guilt and shame. Restitution is something the transgressor does, as opposed to some punishment done to him. 4.1.3 Distributive Justice and Fairness A third type of justice is distributive justice, a concept innately linked to fairness (Rawls 1999, Rischer 2002). Rischer (2002) argues that equity is objective, and therefore not the subjective conjecture of personal tastes. Equity requires that shares be divided impartially, impersonally, and evenly without bias to any claim or demand. Contrary to equity, fairness may change over time and place, so that current 60

PAGE 74

and prevailing norms dictate its rules (Richer 2002). Though there are cultural and social rules that produce the guidelines of fairness, fairness is in a large part individually determined: what is fair to one individual may not be fair to the next. Thus fairness is not equivalent to distributive equity unless there are equal claims. Fairness requires thought, intent, and deliberation, so that distributions meet the needs set out by fairness. Justice, linked to fairness, is an issue of proportion. Whereas equity would give every individual an equal share, justice would give each his due. In distributive justice, equitable distribution is context dependent. Rischer (2002) argues that fairness belongs in justice, but not in economics, as it is an essentially moral issue. He cautions that in game theory and economics, the concept of fairness is skewed to include only the satisfaction of all parties involved. But happiness (satisfaction) is different than fairness pursued as an instrument of justice. "The paramount consideration for fairness as an aspect of justice is not how an individual fares in relation to his own claims but how he fares in relation to the rest of the claimants" (p.16). This statement mirrors the Nash equilibrium and principle of relative fitness maximization. Fairness in the sense of the Nash equilibrium would mean that the situation is 'envy-less,' or that no one wants any other person's share or thinks that their share is deficient with regard to anyone else's share. Though an adequate review of fairness and how individuals define it is impossible here (see Rawls 1999, Walster et al. 1978), fairness is complicated by subjective issues such as the incommensurability of different kinds of goods, and the inability to understand another individual's perspective (historically and experientially defined). Another curious aspect of fairness is that though individuals may publicly agree that equity is important and desirable, it is likely that privately, each individual would prefer more than the publicly-objective fair share (Shroeder et al. 2003). 61

PAGE 75

Finally, fairness is not a concern until competition for a resource is strong (i.e., when demand is above carrying capacity), because until then, each individual may take or have as much as he wants (Schroeder et al. 2003). When resources are limited however, tensions rise with regard to shared distribution, and actions may be taken to punish those who seem to take more than their share, especially if they seem to do so with the intent of exploiting others. Echoes of this in game theoretic situations (Nowak et al. 2000, Rabin 1993) are visible when participants retaliate against free-riding defectors. In ultimatum games punishers seem to prefer the equality of a zero-payoff to the inequitable distribution of monies that, though they produce a non-zero take, favor the exploitative participant. Equity thus seems to prevail in group dilemmas; but if individuals are indeed rational, decision-making should be reduced to a cost-benefit analysis of a given situation. 4.2 Justice and Game Theory In the context of game theory, different forms of justice take on different connotations than when examined in the context of crime (Schroeder et al. 2003). Distributive justice is concerned with the differential payoffs participants receive, thus aligning it with inequality aversion. Retributive justice applies to actions taken by participants to punish defectors, usually based on an emotional reaction to unfair behavior, therefore aligning it with Rabin's (1993) fairness-of-intent theory. Restorative justice encompasses measures taken to compensate participants who are wronged by other participants and represents uncharted territory in experimental economics. (Re)distribution in a game may improve one individual's position (there distributor's) relative to coplayers, or realign all positions through the more fair distribution of payoffs. Punishment may attempt to shape future behavior of a player by dissuading him from defecting in the future. But compensation (restorative justice) in a game will only do restitution for wrongs done by another player. This highlights a major difference in justice in game contexts: both restorative and 62

PAGE 76

retributive paradigms of justice (theoretically) aim at the rehabilitation of a criminal, but in a game situation, restorative justice only compensates the victim while letting the transgressor free-ride. Finally, in social dilemmas and by extension in game situations, individuals concerned with justice may anticipate not only immediate effects but long-term impacts (i.e., private and public resources; social dynamics) (Schroeder et al. 2003). This may be compounded in cultures where anonymous interactions are rare or non existent, because participants cannot suddenly ignore enculturated behavior in the game situation. Thus observed behavior may seem anomalous to economic and evolutionary predictions, but instead make manifest important cultural and social norms like those of fairness and justice that operate with a view to future implications. The justice system employed by Western nations is on the state-controlled retributive end of the justice continuum, while the customary system of PNG has elements of both retribution (usually swiftly employed physical abuse or even death) and restoration (compensation of the victim and his family). From the above discussion, it is clear that a retributive, imprisonment-focused justice paradigm clashes fundamentally with a combination retributive-restorative paradigm, based on swift punishment and victim compensation The imposed retributive judicial system has probably conditioned Papua New Guineans to accept imprisonment as at least a secondary or supporting form of punishment to more swift punition and victim compensation, however. This study aims to elucidate whether either or both retributive or restorative tendencies are visible in economic game behavior. 63

PAGE 77

S. Methods After receiving approval from the Human Subjects Research Committee, University of Colorado at Denver (#2005-083) (See Appendix C), the data for this project were collected over a one-month period in June and July 2005 in the three Au villages of Brugap, Winaluk, and Anguganak. Data collection in each village was completed in one day's time. The villages were within a few hours' walking distance of each other with both Winaluk and Anguganak situated atop the ridgeline. Both Anguganak and Brugap have populations of about 350 individuals and are approximately equidistant from the Station and airstrip. The population ofWinaluk is smaller, at approximately 175 individuals, and Winaluk is farther from the Station than the other two villages so that one must climb up a mountain from the Station and pass through several hamlets of Anguganak Village in order to reach Winaluk. 5.1 Sample Recruitment We selected Brugap, Winaluk, and Anguganak after assessing village willingness to participate in research. We then notified each village via messenger several days prior to the intended date of data collection. Individuals were told that they had the opportunity to volunteer for game-like research, and that they would be paid a nominal show-up fee with the potential to win more money in the research game. Our arrival at each village on the morning of data collection was conveyed by word-of-mouth to any villagers who had chosen to remain in the village on that day 64

PAGE 78

(instead of going to the bush to garden and gather food). We asked all interested men and women, aged 18 and older, to meet us in a central location. As people trickled in, we announced that in order to participate, each person had to attend an initial orientation meeting detailing game play. Latecomers would not be allowed to participate in the action roles of the game, though we did allow some to participate in the inert roles (see below). While the participants gathered, we recorded the name and gender of each person in order to record the number of people in attendance, to attempt to have a gender-balanced group, and to ensure that only those present before the meeting began were allowed to take on decision-making roles in the game. 5.2 The Game The third-party justice game is dynamic33 in that there is a sequential order to player action so that at least some player action is dependent upon previous action by another player (Romp 1997). It is also a normal form game in that the identity of the players is anonymous and all players have complete information about game rules. The original third-party punishment game is a dictator game in that one player determines how to split a sum with another player, with the addition of a third-party who is given an endowment equal to 50% of the endowment allotted to the dictator and the second-party34 The third-party is then given the option to keep his endowment or punish the dictator, at a cost, should the dictator's behavior be deemed selfish or defective. The current game, devised by Tracer, the principal investigator, extends the autonomy of the third-party so that not only may he punish the proposer, he may also compensate the inert second-party, also at an equal cost; or he may both punish and compensate at double the cost. 33 In static games all players act without knowing what other players will do, as in a sealed bid auction or the Prisoner's Dilemma (Romp 1997). 34 For example, i fthe dictator is given an endowment of $10 to divide between himself and the second-party, the third-party is automatically endowed with $5. 65

PAGE 79

More explicitly, Player A (the proposer) divides up a sum of I 0 kina (K 1 0)35 between himself and an anonymous Player B (the second-party). Player A may offer any amount to Player B, from KO to K 10. Player B is inert: B merely receives his payoff at the end of the game after the third-party (anonymous Player C) has the opportunity to sanction Player A or compensate Player B. Player C is given K5 and four options. After hearing Player A's offer to Player B, Player C may Option 1) Do nothing: he takes his K5 and Player A and Player B receive payoffs according to Player A's proposal. Option 2) Player C may punish Player A at a cost to himself; that is, he may pay K 1 (20% of his K5 endowment) to take away K3 from Player A. Both the paid kina (from Player C) and the taken kina (from Player A) return to the experimenter. Option 3) Player C may compensate Player B at a cost to himself; that is, he may pay K1 (20% of his K5 endowment) to add K3 to the endowment of Player B. Again, the paid kina comes back to the experimenter, and the experimenter adds K3 to the payoff of Player B. Option 4) Player C may do both, that is both punish and compensate at twice the cost to himself; he may pay K2 (40% of his K5 endowment) to both take away K3 from Player A and add K3 to the endowment of Player B. The paid kina goes back to the experimenter, and the experimenter transfers K3 from the endowment of Player A to that of Player B. For example: Player A is presented with I 0 kina. Player A offers K3 (30% of K 1 0) to Player B, intending to keep K7 (70%). Player Cis given K5 and told what Player A has offered. Player C then has the option to act or walk away. 35 All game transactions were played using the Papua New Guinean currency of kina. At the time of research, $1 was approximately equal to three kina (K3). The K I 0 sum used in the game is at the upper-end of a day's wage for an unskilled laborer, and is a large sum especially considering most village residents have no wage income (Tracer 2003). 66

PAGE 80

Option I): Player C does nothing. Payoffs: Player A=7 Player B=3 Player C=5. Option 2): Player C pays K 1 to take away K3 from Player A. Payoffs: Player A=4 Player B=3 Player C=4. Option 3): Player C pays K1 to add K3 to Player B. Payoffs: Player A=7 Player B=6 Player C=4. Option 4): Player C pays K2 to do both actions. Payoffs: Player A=4 Player B=6 Player C=3. 5.3 Data Collection We explained the conditions of participation and the rules ofthe game in Melanesian pidgin, using a script that had been back-translated into English in order to check for clarity (See Appendix B). We assured individuals oftheir voluntary participation; that they would receive K2 for taking part in the research but could also earn from KO to K 1 0 during the game; that the research seemed like a game but was really research; and that all game transactions would be anonymous to all other participants. We warned the individuals that the research process would take several hours, and that they were disallowed from talking about the game during this time. Only those aged 18 and over were allowed to participate. Verbal explanation ofthe rules of the game was accompanied by cartoon illustrations, including several examples (See Appendix B). Examples were chosen specifically to show the wide range ofpotential offers and therefore actions allowed in the game. These included offers of K7, K3, K5, and the purely selfish offer of KO. We encouraged individuals to shout out answers to example questions. After the initial briefing as a group, participants were called up individually, alternating between males and females when possible. We interviewed each 67

PAGE 81

individual in order to collect basic personal information (see Appendix B) and each individual received his participation fee at the end of the short interview, before entering the game enclosure (Photograph All). All individuals were designated as Player A until half of the group had finished the game. We then shuffled the Player A data sheets containing offers. The remaining individuals were designated as Player C, acting on randomly matched Player A offers. Players A and C each entered an enclosed space (a room in the community school in Brugap, and small meeting houses in Winaluk and Anguganak) to play the game in the company of the experimenter. On the floor between the participant and the experimenter were three large pieces of paper bearing cartoons to represent each player. Ten K 1 coins were lined up on the paper for Player A; five K 1 coins were lined up on the paper for Player C (Photograph A.l2). Before the game commenced, we checked understanding of the rules of the game, and if necessary used examples for clarification. In both examples and the actual game, the coins were moved to explicitly demonstrate the payoffs for each player. To make his proposal offer, Player A moved any amount of coins from zero to ten onto the paper representing the payoff for Player B. Understanding was clarified one last time to ensure that Player A realized that under the conditions of his offer, the payoffs for both Player A and Player B would be as he dictated. However, Player A was also asked if he realized that Player Chad the opportunity to either punish or compensate by giving up a portion of his endowment. Before Player C entered the room, coins were moved to represent the randomly matched proposal offer Player Chad the opportunity to act on. We then clarified that Player C understood the offer and the proposed payoffs to the other players should Player C choose to abstain from action. In addition, all of Player C' s optional actions were clarified, manipulating the coins to show potential payoffs. After Player C made his decision, the coins were again moved to show the final 68

PAGE 82

payoffs as a result of his action (or lack thereof). We then asked Player C to explain why he chose his particular course of action. Remaining players were assigned to the role of Player B. In the case that there were not enough individuals to complete all trios, we also allowed some late comers who had not been present for the initial orientation session to assume the role of Player B. Because Player B is inert, it was inconsequential if he had limited or incomplete understanding of rules of the game and was absent from the orientation session. Also, due to time constraints and because Player B makes no decisions but merely receives a payoff, Player B was not always interviewed to collect demographic variables, though names, gender, and ages were collected. There were almost always enough latecomers to complete the trios, with most Players B happening upon the gathering, doing nothing except give their name, gender, and age, and walking away with a handsome sum. However in a few cases, one person was assigned the role of Player B for two or more trios. This was usually done when the Player B payoff for a particular round was KO. After all trios had been completed, individuals were called back into the game enclosure to receive their payoffs. A small number of players were asked to listen to one or two hypothetical vignettes about a crime and to choose the appropriate action to amend the crime from a multiple choice bank (See Appendix B). Measures were taken throughout play to ensure anonymity of player decisions, of the identity of specific trios, and of the amounts of payoffs. We distributed payoffs surreptitiously, passing folded bills and coins into the hands of players during goodbye-hand shakes. However, curiosity proved too much for some players as well as for some uninvolved observers who furtively attempted to peer into the room where the game was played, especially during pay-offs. The sample size from all three villages was 133, or 46 trios with five Player Bs participating twice (N=133: Player An =46, Player B n =41, Player C n =46). Tables 2 and 3 summarize the sample by village, role, and gender. Though the total 69

PAGE 83

sample is evenly divided between males and females, when viewed by village there is an imbalance in the number of males and females especially for Winaluk and Anguganak; however, this difference is not statistically significant. On the other hand, even more pronounced is the gender imbalance when viewed by roles. This is inconsequential for Player B, but for the decision-making roles A and B, women outnumber men as Player A 3:2, while men outnumber women as Player C by 5:2. This gender imbalance is statistically significant so that the null hypothesis of no difference between roles by gender may be rejected (chi-square 12.838, p-value 0.002). T bl 2 R I b vll a e o e ,y 1 age an dG d T en er ota s Village Total (N) Player A Player 8 Player C Males Females Brugap 63 2I 2I 21 32 3I Winaluk 29 II 7 II 12 I7 Anguganak 41 14 13 14 23 18 TOTALS N=133 46 41 46 67 66 T bl 3 R I b G d a e oe ,Y en er an d VII 1 age Pia, erA Pia er 8 Pia er C Villa2e N Male Female Male Female Male Female Brugap 63 10 1I 9 I2 13 8 Winaluk 29 2 9 1 6 9 2 Anguganak 41 6 8 6 7 11 3 Totals 133 18 28 16 25 33 13 5.4 Analysis After entering all data into Microsoft Excel, data were imported into SPSS I3.0 for Windows for all statistical analyses. For comparison of the means of groups of nominal data, chi-square tests with Cramer's V measures of association were used. When comparing the means of more than two groups of continuous or nominal data, one-way ANOV A with post-hoc LSD tests were used. All tests were 70

PAGE 84

run with a confidence interval of95% so that the significance level was flagged for p-values of <.05. However, exact significance levels are always provided. For nominal data, cases with missing values were excluded on a test-by-test basis. For continuous data and logistic regression analyses, cases with missing values were excluded on a listwise basis. 5.5 Methodological Issues First, individuals participating as Player A generally had a good grasp of their task in the game after the initial orientation session, and made their decisions and offers quickly. Those participating as Player C, however, often required a review explanation oftheir role and additional examples after entering the enclosure to play the game. This is understandable as actions available to Player C are more complicated. Unfortunately, the difficulty of the Player C task along with experimenter-bias about cognitive ability, may have contributed to gender imbalance. Women tend to be less well-educated than men, as a one-way ANOVA shows statistically (F = 30.036, p-value <0.000 1) (See Figures 1 and 2). Figure 1 Education by Gender 50 ,..40 l! !JO r ""20 10 Male Female Gender Educated Non-educated Figure I: Parentheses indicate percentage of gender, for example, 88% of all males for whom data were available reported at least some education. Figure 2 Education by Gender and Village 71 1 25 g20 .., 15 c 10 r .. 5 Educated Males Educated Females Brugap Winaluk Anguganak Village Figure 2: Labels indicate percentage of gender educated, for example, I 00% of males from Brugap reported at least some education while only 58% of male participants from Winaluk reported any education.

PAGE 85

Participant education is generally predicted by village distance from the Station and thus the community school. Winaluk was the farthest village from the Station and had the lowest education rates: 58% of male participants reported at least some education while only 18% of females reported any education. Our bias may have predisposed us to designate women for the simpler task of Player A or the inert task of Player B while reserving those with at least some education for the role of Player C. Indeed, in Winaluk, where education was much lower than at either Brugap or Anguganak, frustration with a low level of comprehension definitely drove us to choose better-educated individuals for difficult roles. The null hypothesis of no educational difference between villages may be rejected (ANOVA F = 12.464, p value <0.000 1) while the LSD post hoc test showed that while the mean difference between Brugap and Anguganak was not significant (0.569, p-value = 0.389), the mean difference between each of the other villages and Winaluk was significant (Winaluk and Brugap: -2.993, p-value <0.0001; Winaluk and Anguganak: -3.562, p value <0.0001). As discussed in Chapter 6, there is a secondary modal offer of 0%, contrary to previous UG experiments among the Au and elsewhere. Such purely selfish offers seem to corroborate the predictions of economic and evolutionary theory. However, the way in which the game is framed, which Bolton et al. ( 1998) assert may bias game behavior, may have influenced offers. We used carefully chosen language to indicate that the original endowment of K 1 0 belonged to both Players A and B, but that allocation was at the discretion of Player A. Though Hoffman et al. (1994) warn against such a methodology on semantic grounds, citing that this language insinuates that A must give up some amount of the original bank thus driving offers up, the extremely high number of KO offers demonstrates that Players A certainly did not feel any obligation to make a non-selfish offer. Indeed, to counter this obligatory sentiment, one of our verbally and pictorially represented examples used in the orientation session was that of a purely selfish offer of KO. Using this example, and 72

PAGE 86

clearly demonstrating the example with cartoons, gives explicit proof that an offer of 0%, even if punished, ensures that Player A leaves with no less than 70% of the original endowment and that Player B can leave with no more than 30%. The risk of being selfish is thus very tempting, especially if rational proposers presume the rationality of other players so that no Player C would altruistically punish or compensate. It is questionable if the use of an extremely clear example offer of 0% (among other more altruistic offers) encouraged participants to offer less than they normally would had they been left to figure out the payoffs on their own, without the help of tangible, pictorial representations. Finally, the Player C role is very difficult, requiring considerable explanation. To prevent wasting time explaining the rules of the game to each player individually, we chose to explain the game first to all players in a pre-game orientation. Though they were discouraged from talking about the game, this may have allowed collusion about game-decisions as players waited several hours for their tum to participate in the game. This may also explain the high rate of KO offers, previously absent in other economic experiments among the Au (Tracer 2003, 2004). 73

PAGE 87

6. Results and Discussion 6.1 Quantitative Results 6.1. 1 Frequencies of Variables A total of N = 133 volunteers from the three different villages of Brugap, Winaluk, and Anguganak participated in the third-party justice game. The sample included 67 males and 66 females ranging in age from 18 to 80. Table 4 summarizes the frequencies of scale variables collected during qualitative interviews. Table 4 Freguencies of Variables Variable n Minimum Maximum Mode Mean S.d. Age (yrs) 85 18 80 26 33.13 13.33 # of children 119 0 12 3 2.27 2.48 Education (grade) 118 0 10 0 3.65 3.37 # of gardens* 115 0 1 I 5 4.63 2.19 Church attendance** 118 0 4 4 2.73 1.69 Cash crop income*** 103 0 1500 100 129.50 240.73 Work income**** 119 0 400 0 15.01 61.30 NOTE: Due to time constraints, all variables were not collected for every participant, particularly for inert Player B. In addition, some participants either did not understand questions, or did not know the answer to questions (e.g., did not know their age), so that there are missing data. Number of gaden kaikai, or food gardens. **Church attendance denotes number of times of attendance per month, assuming that one month has four Sundays so that the maximum answer is four. ***Cash crop income in Kina per month, usually from selling cocoa, coffee, and/or vanilla. ****Wages earned by working in Kina per month. No paying jobs exist in the villages, so that all participants earn their income at local schools, the clinic (nurses), or at the Station. In fact, of the 115 participants for whom data were available, only 14 (11.8%) are wage workers. Table 5 presents the results of one-way A NOVA comparison of means of all variables across the three villages. Though several variables differed significantly 74

PAGE 88

between villages (See Figures 3 and 4 for distributions of significantly different means; Figure 3 shows the continuous data associated with the variables of education, church attendance, number of gardens, and cash crop income, while Figure 4 shows the nominal data associated with existence of vanilla and cocoa gardens as a source of cash crop income), ANOV A analyses of Player A offers and Player C actions show that we cannot reject the null hypothesis of no difference between villages for game data (Player A offer: F = .439, p-value = .645; Player C action: F = .446, p-value = .641 ). Therefore, post hoc multiple comparisons between the villages are not reported for the categories that were significantly different, and game data from the three villages are lumped into one sample for further analysis. Table 5 ANOVA Comparison of Variable Means Between Villages Variable F Gender Age (years) Marital Status # ofWives #of Kids Education Church Attendance #of Gardens Coffee* Cacao* Vanilla* Cash Crop Income/Mo. Work (yes/no) Work Income/Mo. Player A Offer Player C Action .732 .020 .432 2.646 1.337 12.464 5.666 4.238 .059 3.123 3.682 9.763 .116 2.270 .439 .446 p-value .483 .981 .650 .080 .267 <.0001 .005 .017 .943 .048 .028 <.0001 .891 .108 .645 .641 Note: Variables with significant differences between villages (p-values < .05) are in bold. *Coffee, Cacao, and Vanilla are nominal categories wherein participants answered yes or no to cultivating these cash crops. 75

PAGE 89

Figure 3 Distribution of Selected Means Figure 4 Vanilla and Cacao by Village 25-t------20 +---------1----l Brugap Winaluk 10 C Anguganak Education No. of Cash Crop Attendance Gardens Income Variable Figure 3: Cash crop income is I /10 of actual income. 6.1.2 Player A Offers 45 40 35 25 r20 .. 15 10 Brugap Winaluk Village Anguganak Figure 4: Labels indicate percentage of participants who have cash crop income from either vanilla or cacao; for example, 70% of all Brugap participants grow and sell at least some vanilla, while almost all (97%) of participants from Winaluk have cash crop income from vanilla. Table 6 summarizes the distribution of Player A offers (n = 46) in the third party justice game experiment. Offers ranged from KO to K9. Offers produce a multi modal distribution, with primary modal offers at K3 and K4, each representing 19.6% of total offers (together making up 39.2% of all offers). Surprisingly, a strong secondary mode exists at offers of KO, totaling 15.2% of all offers. The mean offer was K3.30 with a standard deviation of2.38. 76

PAGE 90

Table 6 Frequency of Player A Offer (Kina) at Brugap, Winaluk, and Anguganak Offer n Percentage 0 7 15.2 I 5 I0.9 2 4 8.7 3 9 19.6 4 9 19.6 5 6 13.0 6 1 2.2 7 2 4.3 8 I 2.2 9 2 4.3 10 0 0.0 Total 46 IOO.O Mean offer= 3.30, s.d. = 2.375. Modal offers are in bold. Primary modes= 3 and 4; secondary mode= 0. The mean offer was 3.33 (s.d. 2.59) for males (n = I8, 39.I %) and 3.29 (s.d. = 2.28) for females (n = 28, 60.9%). A two-tailed t-test reveals that there is no significant difference between these (t = .066; p-value = .948). Of the other individual variables, only cash-crop income (Pearson correlation= -.377, p-value = .013) proved a good predictor of Player A offers. Inversely correlated, the more cash crop income participants reported, the lower the Player A offer. Regression, correlation, ANOVA, and univariate analyses show that no other single variable or combination of variables is a good predictor of Player A offers. 6.1.3 Sanctions and Compensations Table 7 and Figure 5 show the distributions of Player A offers along with Player C actions to either punish Player A, compensate Player B, or do both (n = I6 action-takers out of n = 46 Player C). More than a third (34.8%) of all Players C took some action, all for offers of KO to K4, so that no Player C took any action for any 77

PAGE 91

fair (K5) or better (>K5) offer. Astonishingly, Player C punished and/or compensated at or above half the time for offers of KO to K3; Player C also punished 22.2% of the time for offers of K4. In other words, when faced with selfish offers of0%-30%, the majority of Players C took action. Table 7 Frequency of Player A Offer (Kina) and Player C Action at Brugap,Winaluk, and Anguganak %Action %of Offer nojji-r naction per noffer Total Action 0 7 4 57.1 25.0 1 5 3 60.0 18.8 2 4 2 50.0 12.5 3 9 5 55.6 31.2 4 9 2 22.2 12.5 5 6 0 6 1 0 7 2 0 8 1 0 9 2 0 10 0 0 Total 46 16 100.0 Figure 5 Distribution of Player A Offer (Kina) and Player C Action 9 8 A-----7 6 5 Frequency 4 3 2 1 0 0 2 3 4 5 6 7 8 9 10 Offer and Action F1ayer A Offer F1ayer C Action Taken Player C acted on offers ofKO to K3 about half the time and on offers ofK4 about a quarter of the time. No Players C acted on offers ofK5 and above. 78

PAGE 92

Table 8 and Figure 6 itemize Player C action with offers of KO through K5 while Table 9 shows the distribution of Player C action as a whole and by gender. Punishments occurred for a wider range of offers (KO to K4) than did compensations (KO to K3); both occurred over wider ranges than the 'do both' action (KO to K2). The overall sample is well-balanced for gender (67 males and 66 females); however, the Player A sub-sample is biased toward a female majority while the Player C sample is biased toward a heavy male majority: 71.7% of Player C participants were male, 28.3% were female. Overall, the majority (65.2%) of Players C chose to do nothing and take the K5 payoff. However, 34.7% of the time Player C chose to punish, compensate, or do both; more specifically, 30.3% of males and an incredible 46.2% of females did so. These rates of Player 3 action are extremely high, especially considering that 26.1% of offers were K5 or above ('fair' or hyper fair). The astonishingly high rate of altruistic punishment and compensation of anonymous trios stands in stark contrast with economic theory that predicts Player C should never act and thereby sacrifice a portion of his payoff. As discussed below however, punishment may be congruent with the evolutionary theory of relative fitness (or utility) maximization, especially as the tendency to punish may be less a reaction to fairness and more to do with even payoff distribution. Altruistic compensation is a bit more difficult to explain theoretically, even if the motivation is toward even payoff distribution. Table 8 Player C Action by Offer (truncated at K5 because no action was taken for offers of K5 or greater) Offer No Action Punish Compensate Do Both 0 3 1 1 2 1 2 0 2 2 2 1 0 1 3 4 2 3 0 4 7 2 0 0 5 6 0 0 0 79

PAGE 93

Figure 6 Distribution of Player C Action for Offers 0%-50% 10 9 c 8 0 :g 7 0 Do Both 6 .... 0 Compensate 0 >. 5 u Punish c 4 G) :I 3 No Action C" 2 LL 0 0 2 3 4 5 Offers Table 9 C Decisions Overall and Gender n Percentage Male(%) Female (%) No Action 30 65.2 23 (69.7) 7 (53.8) Punish 6 13.0 5 (15.2) (7.7) Compensate 6 13.0 4 ( 12.1) 2 (15.4) Do Both 4 8.8 (3.0) 3 (23.1) Total 46 100.0 33 (100.0) 13 (100.0) Notes: --Player C could only punish Player A and only compensate Player B. To punish, Player C gave up Kl (of his K5 allotment) so that the experimenter would take away K3 from Player A; to compensate, Player C gave up K I so that the experimenter would add K3 to Player B. --To both punish (Player A) and compensate (Player B), Player C gave up K2 (leaving her with K3) in order to take away and add K3 to Players A and B, respectively. --Percentages in parentheses are percent within gender. --Of the total Player C sample, males made up 71.7% and females made up 28.3%. 80

PAGE 94

Though the sample size was too small to stratify and statistically compare Player C action by any variable36 it is interesting to note that nearly half of all women (6 of 13) took action whereas less than a third (10 of33) of all men chose to punish, compensate, or do both. Whereas an extraordinary 23.1% of women chose to sacrifice K2 in order to both punish and compensate, only 3.0% of men ( 1 of 33) chose to do so. Also interesting is the fact that the male sample skews toward punishment while the female contingency skews toward compensation or the punishment-compensation combination. Further research with a well-balanced gender sample will help elucidate whether gender statistically influences the tendency to punish, compensate, or do both, and will potentially augment a growing body of research on gender tendencies in economic games. Previous experimental results suggest that women are less-selfish than men (Croson and Buchan 1999, Eckel and Grossman 1998), at least when altruism is expensive (Andreoni and Vesterlund 2001). Our results similarly suggest that women have a higher tendency than men to altruistically act in the third-party game, usually to 'do both,' or compensate, rather than punish. Moreover, if 'doing both' may be considered a proxy of the desire for even distribution, the results may 1) further corroborate previous research that shows that while men tend to be either perfectly selfish or perfectly self-less, women tend to share evenly (Andreoni and Vesterlund 2001); and 2) may prompt debate about theories arguing that punishment is a reaction to unfairness and means towardjustice. 36 When Player C action is divided up into four categories (no action, punish, compensate, punish and compensate), 75.0% of cells have expected counts less than 5. Even with Player C data collapsed into two categories (no action and action), 25% of cells still have expected counts less than five. 81

PAGE 95

6.2 Qualitative Results 6.2.1 Interviews and Vignettes Post-game interviews with randomly selected third-parties demonstrate that players were motivated by a variety of sentiments when deciding to punish, compensate, do both, or walk away. Of those that walked away, some were apathetic ("let Player A and Player B work it out;" "it's none of my business") while others were conflicted. The latter wanted to even up the pay-outs between all players, but were either 1) unwilling to give up money to do so or 2) without recourse, that is, unable to add or subtract from either player's payoff to reach equality (K5:K5:K5) (for example, when faced with hyper-fair offers since there was no option to reduce the payoff of Player B). The majority of punishers and those that both punished and compensated cited the desire to even up distribution of payoffs to equality when asked about their decisions. The compensators, however, gave more emotional answers, saying "I feel sorry for Player B," or "I should help my brother/sister." Equitable distribution as a motive for third-party action will more fully be discussed below. In addition, a very small sub-sample of third-party participants (n = 11.2%) were asked to listen to one or two vignettes about crime, and describe the appropriate punishment (See Appendix B for vignettes and See Figure 7). Sixty-three percent stated that compensation is the appropriate punishment for petty crime (Vignette 1 ), at least as a primary punitive measure in conjunction with imprisonment should the guilty party refuse to pay compensation. Imprisonment alone was satisfactory for only 9% of the sample, while imprisonment coupled with a beating was the choice of 27% of the sample. Responses to vignettes about more serious and violent crime varied with respect to imprisonment vs. the death penalty, but always included compensation to the victim's family or clan. Qualitative data thus support the assertion that, probably as an extension of the reciprocal exchange system that values 82

PAGE 96

generosity, compensation as a vehicle for restorative justice is an integral part of punition. Figure 7 Vignette Responses 4+-------3.5 +--.......... "7'nT--->3 u i 2.5 2 I.L 1.5 0.5 0 Compensate Compensate Conditional Imprisonment Imprisonment + Beating Appropriate Punitive Measure Figure 6: Compensate Conditional refers to the responses for which compensation was the first choice of retribution if the transgressor could be persuaded to do so. If the transgressor refused, either imprisonment or a beating would be appropriate as a secondary course of punitive action. 6.3 Discussion Table 10 summarizes offers from previously published ultimatum game experiments. The range of UG and DG offers from games performed in small-scale societies (Henrich 2000, Henrich et al. 2001) vary much more than those of UGs played with university students in a diverse group of industrialized countries (Pittsburgh; Ljubljana, Slovenia; Jerusalem; and Tokyo) (Roth et al. 1990); and in Yogyakarta (Indonesia) (Cameron 1995). Whereas the mean student offer was between 43% and 48% and the mode consistently 50%, mean offers from the small scale societies ranged from 26% to 58%, with a modal range of 15% to 50% (Henrich et al. 2001 ). Results from previous and current studies in Papua New 83

PAGE 97

Guinea are between the student and small-scale society offers. Among the groups from Papua New Guinea, the mean reported here is lower than previous results at 33%, but the modes are consistent with previous findings. Table 10 Summary of Mean and Modal Offers in Ultimatum Games Mean Mode* University students 43%-48% 50% Small-scale societies2 26%-58% 15%-50% Au (ofPNG) 3 43% 30% Gnau (of PNG) 3 38% 40% Current results 33% 30%, 40% *Hyphens indicate range. Commas separate bimodal offers. 1(Roth et al. 1990, Cameron 1995) 2(Henrich et al. 2001) 3(Tracer 2003) The following are a reiteration of predictions followed by discussion of each. I. The proposer will never make a non-zero offer. If individuals are selfish, as economic and evolutionary theory predicts, offers in the third-party justice game should not exceed 0%, especially under conditions of anonymity. Results show primary modes at offers of30% and 40% (K3 and K4), which is consistent with offers found in previous ultimatum games performed in the same area of Papua New Guinea (Tracer 2003). The primary modes are in direct opposition to Hypothesis I and therefore oppose the predictions of evolutionary and economic theory. However, results also show a strong secondary mode at 'perfectly-selfish' offers of KO. Tracer (2003) recorded no offers of 0% in a UG previously played among the Au. The high rate of zero offers corroborates Hypothesis I, and therefore seems to corroborate economic and evolutionary predictions that individuals are selfish. Two 84

PAGE 98

considerations may otherwise help explain the high rate of zero offers, one theoretical and the other methodological. First, should economic theory be correct in assuming that all players are rational maximizers of utility, these rational individuals should presume that their peers are likewise rational. Thus, Player A would deduce that no Player C would punish a low offer because it would decrease Player C's utility. Alternatively, this rationality might explain selfish offers in a different way. In the UG previously performed among the Au by Tracer (2003), the proposer may reasonably expect punishment from the one he wrongs so that offers are non-selfish. In the third-party justice game however, the threat of punishment is less severe because proposer-behavior does not affect third-party payoffs. This lack of perceived threat could encourage more selfish offers than have previously been observed. A second methodological possibility as discussed in Chapter 5, is that during the game orientation session, several verbal and pictorial examples were given to illustrate potential payoffs according to game decisions37 One ofthese, among others, was the offer of KO. With the cartoon representations, it was very easy to see that in the worst case scenario of Player C punishing Player A for a purely selfish offer of zero, Player A would leave the game with no less than 70% of the original stake (K7). It is possible that even with good understanding of game rules, many players would not have so clearly envisioned their winnings for a KO offer without the pictorial example. Hyper-low (purely selfish offers of KO) offers stand in stark contrast to hyperfair offers, or those greater than 50% of the original sum. 13,0% of all offers were hyper-fair (>50%), the existence of which alongside modal offers of30% and 40% clearly opposes Hypothesis I. However, this finding is consistent with a previous ultimatum game performed among the Au where 14.5% of offers were hyper-fair (Tracer 2003). Tracer (2003) argues that the existence (and rejection) of 37 Bolton et al. 1998 emphasize the importance of the game frame on offers. 85

PAGE 99

hyper-fair offers in the ultimatum game (rejected >50% of the time) may be explained by the previously described cultural premium placed on generosity, while rejections may further be explained by the reluctance to be indebted to anyone that should give too-substantial a gift. Unfamiliarity with anonymous exchanges also helps to explain why, even when assured that identities of players would only be known to the experimenter, both offers and rejection rates of 'fair' offers are high. Even so, extreme, hyper-fair offers of90% are anomalously over-generous. A further, unlikely possibility might be experimenter-influenced costly signaling (Henrich 2000, Hoffman et al. 1994, Hoffman et al. 1996); or that the participants, who knew the primary investigator very well (after 17-plus years of interaction during the course of various anthropological studies), wanted to appear non-selfish to the experimenter for some ulterior motive. This may have been further impacted by the fact that the primary investigator is known to give villagers his remaining supplies (kerosene lamps, food, bedding, mosquito nets, flashlights, etc.) at the end of the field season3!l. Hoffman et al. 1994 argue that not only may experimenter influenced costly signaling drive up Player A offers, it could drive up the rate of altruistic compensation. Indeed, the rate of Player C action in the current third-party punishment experiment at 34.7% is slightly higher than the punishment rate of 32.8% that Tracer (2003) found in an ultimatum game experiment. However, rates of hyper fair offers and ofthird-party action were evenly distributed between villages. For experimenter-bias to be at work we might have expected that hyper-generosity be more prevalent in Anguganak, the village where the experimenter resides when he does his fieldwork; and the village in which most of the post-field-season supplies are doled out. Nonetheless, high rates of altruistic punishment and compensation, as well as the existence of non-selfish and hyper-fair offers are in direct opposition to the economic assumption that individuals are selfish, absolute utility maximizers. JK One participant admitted to the primary investigator that he had lied during his interview about church attendance in order to look good for the author in pre-game interviews. 86

PAGE 100

II. Third-party players will never act because it is costly to do so. The third-party payoff, 50% of the endowment available for division between the proposer and inert recipient, is allotted independent of proposer and recipient behavior. According to both economic and evolutionary theory, the third-party should never punish or compensate because it costs him 20% to do so without any potential gain. The third-party should certainly never both punish and compensate because it costs 40% of his endowment. These assertions and the above hypothesis are clearly refuted by empirical results. Overall, a non-trivial 34.8% of third-party participants punished, compensated, or did both. And when faced with unfair offers of 0%-30%, the majority of third-parties acted. Punishment and compensation are equally costly (K 1 out of K5, or a 20% sacrifice). One ofthe aims of this study is to elucidate whether tendencies based on retributive or restorative justice play a role in game behavior. In a setting where the retributive justice system predominates, punishment would be expected to prevail over compensation. PNG has historically relied on both swift punition (retributive justice) and victim compensation (restorative justice), so that we might expect to find a mixture of punishment and compensation in third-party behavior. Overall, players punished and compensated with equal frequency: 13.0% punished and 13.0% compensated, so that more than a quarter of third-parties (26.0%) punished or compensated. Third-parties punished or compensated more often than they 'did both' actions (8.7% ofthe time). When viewing third-party action by gender however, more interesting trends emerge. Females compensated more than they punished: 15% of females compensated Player B and 7% of females punished Player A. Astoundingly however, females both punished and compensated at a greater rate than they did either action; 23% chose to do both actions, incurring a 40% cost to do so. Males punished more often than they compensated, and did each action more often than they both punished and 87

PAGE 101

compensated: 15% of males punished, 12% compensated, and 3% did both actions. Though the sample size was not large enough to show a statistical difference in gender-related behavior, further study with a larger and gender-balanced sample will help determine if gender indeed is statistically correlated with third-party behavior and the tendency to punish or compensate. Altruistic punishment in ultimatum games-rejecting non-zero offersconflicts with canonical evolutionary and economic theory and has been argued to be an emotionally driven reaction to unfair (or selfish) behavior (Bolton et al. 1998, Bowles and Gintis 2002a, 2002b, Fehr and Gachter 2000b, Fehr et al. 2002, Macintyre 2004, Rabin 1993). Though resentment ofunfaimess is arguably universal, the definition of what is fair differs culturally (Henrich 2000, Henrich et al. 200 I), as supported by the above discussion of reciprocal norms in Papua New Guinea. Altruistic punishment in the third-party justice game also seems at first to conflict with both standard evolutionary and economic theory as well. But while punishment is contrary to (economic) predictions of individuals as absolute utility maximizers-utterly self-interested individuals who will maximize utility regardless of the utility of other individuals in the population-it may be explained in the context of relative utility maximization (Tracer 2003). Individuals evaluate their own fitness in reference to the fitness of others and choose the best strategy under the given conditions in order to maximize their own fitness. So by sacrificing 20% of his utility in the third-party game, a punisher decreases the utility of Player A by 30% or more39 For example, ifPlayer A proposes an offer ofKO, the proposed payoffs are Player A: 10 Player B: 0 Player C: 5. Should Player C decide to punish, the payoffs become 39 K3 punishment of a KIO pay-out is 30%; K3 punishment of a K9 pay-out is 33.3%; K3 punishment of a K8 pay-out is 3 7 .5%; etc .. 88

PAGE 102

Player A: 7 Player B: 0 Player C: 4. Relative to the original Player A to Player C ratio of 10:5, the after-punishment ratio of7:4 is much more favorable to Player C. Thus what seems like altruistic punishment is actually self-interested punishment as a means to produce a more even distribution of utility (at least where Players A and Care concerned). More difficult to explain, however, is the compensation of Player B. Luck alone determines the payoff for inert Player B; he must rely on the benevolence of either Player A or Player C for any pay-out, much less a large one. But Player C should never step in to boost the payoff of any player as it contradicts both absolute and relative fitness. Doing so may be evidence of pure altruism; though theories of unconditional altruism do not entail punishment, they may explain why players compensate. More likely however, are theoretical explanations about fairness and justice. Fairness theories may further help to explain behavior in the third-party justice game. Though Rabin's (1993) theory offairness-of-intent may help explain rejections (altruistic punishment) in ultimatum games, it fails to explain third-party behavior here because proposer-action does not affect the payoff of the third-party. Models of equity and inequality aversion (Bolton and Ockenfels 2000, Fehr and Schmidt 1999 respectively) do give insight into the motivations observed in the third-party justice game. Though it is readily plausible that inequality-averse third parties might take away money (i.e., punish) from co-players to make payoffs more equitable, it seems less feasible that they might add to the payoffs of co-players (i.e., compensate) at a cost to self just to distribute payoffs more evenly. However, further examination of qualitative results adds to the explanatory strength of these models. Though no Player C took action for offers at or above equity (K5-K10), quantitative data are misleading. When asked to consider a hyper-fair (>50%) offer, Players C frequently expressed dissatisfaction for the inability to take away money 89

PAGE 103

from Player B or add money to the payoff of Player A 40 This suggests that more than a drive to punish or compensate, Player C wanted to ensure that payoffs were evenly, or close to evenly, distributed between all three players, even though this required taking a loss of 20%-40%. Thus Players C seem to be concerned with the status of their payoff relative to the other players, not just in maximizing self-utility (by reducing the endowment of players with larger endowments) but also in maximizing the utility of 'poorer' co-players (by compensating Player B, and asking if it is possible to compensate hyper-fair Player A). Fehr and Schmidt (1999) assert that individuals who are inequality-averse will make sacrifices to both decrease the endowment of the better-off and increase the endowment ofthe worse-off. This approach requires a refinement of the traditional economic view of absolute utility towards one that acknowledges an individual's recognition of self-utility with reference to that of peers (i.e., individuals are both selfand other-regarding), not unlike the evolutionary idea of relative fitness emphasized by Tracer (2003). Moreover, it may also suggest that a sense of distributive justice drives game behavior, with the goal of fair distribution of material gains. If indeed Au players utilize both compensation and punishment as a means to facilitate even distribution of utility, because of inequality aversion, cultural proclivities, or sense of justice, it highlights the importance of both altruistic punishment and altruistic rewards for the maintenance of cooperation (Fehr and Rockenbach 2003). In addition to inequality aversion, culture may also influence third-party behavior. The proliferation and strength of reciprocal relationships not unlike 40 For example, one third-party who understood the rules of the game very well prior to entering the game enclosure (unlike many), came in and immediately stated that he knew exactly what he was going to do: take the K5 allotted to him and walk away. However, when he was told that the randomly matched Player A offer to which he might respond was an astoundingly hyper-generous K9, he was at first shocked at the generosity, and then determined to reward Player A for his hyper generosity by adding to his small K I payoff. When told that he could not add to the payoff of the proposer (Player A), he wanted to take money away from the K9 payoff of Player B. When told that he could not do that either, he left the game-room as if defeated, upset that he could not more evenly distribute the payoffs from his game. 90

PAGE 104

Trivers' reciprocal altruism may influence third-party action so that village-mates playing the game receive a fair, or fairer than proposed, share. This may be compounded by both the high degree of relatedness of villagers, as well as the lack of anonymous interactions in everyday life. Individuals might imagine that their anonymous co-players could very well be a child, sister, parent, friend, or clan member who might be able to return the favor in the future. Or, in a culture where anonymous interactions do not exist, players may be unable to trust experimenter claims that the game is a truly anonymous exchange, and assume that, should player identities be found out, game behavior could impact future interactions. This could also generate generosity of offers as well as promote compensation in order to both self-protect and ensure social bonds. Finally, customary practices for the treatment of crime may produce the motivation to compensate in the game. Stemming from reciprocal norms, the traditional justice system in Papua New Guinea is both restorative and retributive, while the formal system is decidedly retributive. In other words, paying compensation to victims and their families or clans makes up part or all of the punishment of transgressors. In fact, the enactment of the Compensation Act of 1991 was a formal attempt to marry the restorative, customary justice system and the Western, retributive system that has been imposed since the early 201h century under colonial rule. It is likely that a sense of justice based on the customary system of restoration influences game behavior to promote compensation by third-parties. However, these third-parties punished and compensated at precisely the same rates, while a few players chose to both punish and compensate. These results suggest that both the retributive and restorative systems are equally important. Indeed, the responses to hypothetical vignettes further corroborate the idea that a combination of retributive and restorative justice is favored. However, the higher female tendency to compensate or both compensate and punish is very interesting. Though inconclusive because of such a small sample size, this may suggest that women are less selfish 91

PAGE 105

than men; that they are more inequality averse; and/or that they may favor a combination of restorative and retributive justice over pure restorative justice, but restorative justice over pure retributive justice. This stands in opposition to the tendency of men to favor retributive justice over restorative justice, but restorative justice over a combination of the two. Perhaps this difference is in fact explained by education, as seen through gender. As discussed in Chapter 5, men are better educated than women. Exposure to Western ideas and the retributive (based on imprisonment) justice system that formally presides over PNG may prompt men to favor punishment over restoration. Again, a large, gender-balanced sample that controls for education might elucidate whether this possibility is a reality. 92

PAGE 106

7. Conclusion and Future Directions 7.1 Conclusion Ubiquitous social norms shape and constrain human behavior. Reciprocity, both positive and negative, has been argued to be a key enforcement mechanism of the social norm of altruistic cooperation (Fehr and Gachter 1998). A refinement of this statement is that while negative reciprocity (altruistic punishment) enforces the social norm of altruism, positive reciprocity reinforces this social norm. Neither standard evolutionary theory nor game theory built on neoclassical economic theory can fully explain social preferences for altruistic cooperation and punishment, especially in one-shot encounters. A growing body of empirical evidence at times supporting and at times conflicting with existing theory produces a confusing mass of overlapping descriptions and explanations. The results of this third-party justice experiment are likewise difficult to reconcile using pure evolutionary or economic theory, as results both support and contradict the predictions these theories offer. While offers of 0% in the third-party justice game among the Au seem to support the selfishness axiom of such theory, the bimodal offers at 30% and 40%, as well as hyper-fair offers, utterly contradict it. However, both purely selfish and hyper-fair offers could be explained by methodological issues of game frame (e.g., player collusion, the pictorial example of a 0% offer, and lack of experimenter anonymity). In addition, cultural ideas about reciprocity and the lack of anonymous real-life interactions may also play a role in producing hyper-fair offers. Threat of 93

PAGE 107

punishment or local ideas about fairness (i.e., generosity) mostly likely explain the modal offers. The high rate of altruistic punishment and compensation at 34.7% also contradicts standard economic and evolutionary theory. Though punishment may be explained in terms of relative fitness maximization, fairness theory, or inequality aversion, compensation is more difficult to reconcile. Theories of inequality aversion and distributive equality may help to explain compensation in the game, but again, culture also probably plays a role in third-party compensation as reciprocal norms emphasize generosity and the traditional justice system places a high premium on restoration ofwrongs to victims through compensation. Most variables did not predispose tendencies in proposal or third-party action, with the exceptions of 1) the inverse relationship of cash-crop income and offers, and 2) provisional (but statistically insignificant) evidence that gender impacts the tendency to punish, compensate, or do both. Females tend to compensate or do both more than do males. Overall however, both punishment and compensation were chosen with equal frequency, potentially highlighting social influences from the justice system made manifest in game behavior. Indeed qualitative results suggest that participants believe that a mix of retributive and restorative justice best addresses both petty and violent crime. This is likely derived from the modem PNG justice system, a complicated combination of the customary practices of victim compensation, customary practices of swift physical punishment, as well as the colonization-imposed retributive system based on imprisonment. 7.1.1 On the Evolution of Altruistic Cooperation For a complete picture of the evolution of altruistic cooperation, all of the main theories iterated in Chapter II are necessary. Altruism likely began amongst very close, genetically related kin, which, whether rationalized or not, augmented the inclusive fitness of the kin-group. As cognitive processes evolved, group size 94

PAGE 108

increased, and the genetic relatedness of individuals decreased, humans extrapolated cooperative tendencies from kin to the new group of pseudokin (Macintyre 2004) with whom interactions were repeated and frequent. If cooperation was reciprocally returned with cooperation, so be it. But if cooperation was met with defection, the individuals who could discern the identities of the defectors or who could remember past experiences were selected for as they could reciprocate defection with defection. Thus tit-for-tat, or what Trivers (1971) called reciprocal altruism, was born. But as groups continued to grow in size as did the number of partners with whom one could interact, more efficient mechanisms for communicating and remembering the identity of cooperators had to evolve, making the way for indirect reciprocity and costly signaling. An individual able to discern 'honest' signals and ignore 'cheap talk' was favored by natural selection. Finally, as group size grew so large that anonymous and one-shot interactions grew frequent-thereby eliminating the efficacy of signaling and the ability to base decisions on previous experience-a stauncher mode of enforcement was necessary to deter defection. Altruistic punishment, then, is not a theory that stands alone to explain the evolution of cooperation, nor does it go beyond explication of what is necessary to maintain cooperation among large groups. Altruistic punishment is a means toward maintaining cooperation, reciprocal or otherwise. It is a social norm enforcer. The results presented here show that altruistic compensation may also act as a social norm reinforcer. 7.2 Future Directions All players knew that the other members of their trio were fellow members of their village. Bowles and Giotis (2002b) argue that the motivation to punish is stronger when the identification of the group is known (even if individuals are anonymous), so that strong reciprocity is stronger when group stability is high. By extension, we might assume that the motivation to altruistically reward (compensate) 95

PAGE 109

may also be stronger when group stability is high. The high degree of inter relatedness ofvillage members, either genetically or reciprocally, ostensibly makes it difficult for participants to grasp the idea of an anonymous interaction and the fact that no future interaction will depend on behavior confined to the game. If trios were made up of members of different villages or even different language groups with whom interaction is minimal or non-existent, it would reduce or eliminate the importance of such confounders, namely the inter-relatedness of participants and anonymity, thus allowing issues of justice, fairness, and inequality aversion to come to the forefront. It would also be interesting to give the third-party the option to take action on hyper-fair offers. So, in addition to punishment of the proposer and compensation of the inert second-party, the third-party would also be able to 'punish' inert Player B for (the good luck of) receiving too-high a payoff and 'compensate' Player A for his (hyper-) generosity. This type of justice game will help elucidate whether inequality aversion based on distributive justice (fair distribution ofpayoffs) drives player behavior, or if it is indeed restorative retributive justice that impacts third-party action. Finally, as mentioned above, further testing with a large, gender-balanced sample is necessary to understand if gender influences third-party action. Female participants in this study compensated more often than men, suggesting they are more prosocial than men. This provisionally corroborates similar findings from a dictator (Eckel and Grossman's 1998) and trust game (Croson and Buchan's 1999) that conclude that women are more selfless than men. A possibility here might be that because men in Papua New Guinea are better educated than women, they might have more exposure to Western ideas including those about retributive justice. A study controlling for education may be helpful in illuminating this possibility as well. 96

PAGE 110

7.3 Concluding Remarks In summary, players in the third-party justice game have heterogeneous preferences, both in offers and in third-party actions. Unlike many ultimatum and dictator game experiments performed in geographically and culturally diverse locales, a non-trivial percentage of hyper-fair offers, as well as a large amount of purely selfish (0%) offers exist. Only cash-crop income is a statistically significant predictor of offers. Gender may predict the tendency towards compensation over punishment, though a larger sample size is necessary to test this assertion. Though results show that the third-party only punishes and compensates for offers less than equality (
PAGE 111

APPENDIX A Map and Photographs Map A.l P apua New Guinea I ... M'JtnHP.AClFIC '!i. 0 0 100 200 .300 mi OCEAN .. .._._ ........ ....._l:_.'f2!?r _ ... __ .Jo Coral Sea Map courtesy of CIA World Factbook (htt.p://www.odci.gov/cia!publications/factbook/geos/pp.html) Arrow indicates the approximate location of Anguganak. A traditional-style house. Anguganak, PNG. 98 A more modem, woven-walled house atop stilts. Bogasep, PNG.

PAGE 112

An even more modem house with corrugated roof. Anguganak, PNG. Photograph A.4 Mama Opa scraping sago. Anguganak, PNG. Sago scraping with a bamboo hammer. Anguganak, PNG. 99 Photograph A.6 Meini washing sago, using her hand-made bamboo processor. Anguganak, PNG.

PAGE 113

A typical meal of sago jelly (left) and greens (right). The greens here were stewed with tinned fish. Anguganak, PNG. Community school (left). Anguganak, PNG. The author conducting pre-game interviews. Game enclosure (right). Winaluk, PNG. Photo courtesy of David Tracer. 100 Market day at the Station. In the background are the hangar (left) and trade-store (right). Angugnak, PNG. Sarah grooming Janet. Janet pinching lice. Angugnak, PNG. The principal investigator, David Tracer, conducting the game with a third-party participant. The three pieces of paper hold the amount of money in coins allotted to each player. Brugap, PNG.

PAGE 114

APPENDIXB Script, Data Sheet (Pre-game Interview), Post-game Interview and Vignettes 1.1 Introductory Comments 1. Tenkyu olgeta long kam long hia tude. Dispela wok (em i olsem liklik pilai o gem) em bai kisim longpela taim liklik kain olsem sikispela haua samting olsem na sapos yu no inap stap long dispela longpela taim, yu mas tokaut nau. Pastaim bipo mi wokim dispela wok, mii laik givim yu liklik toksave long wanem kain saating mi mekim na wane mol lo na pasin yu mas bihainim long wokim dispela wok. Translation: Thank you all for coming here today. This research (work) (which is like a little game) will take a pretty long time to play-maybe six hours or so-so if you cannot stay such a long time, please say so now. Before we begin, I'd like to give you a little talk about what we'll be doing today and the things you'll have to know before participating. 2. Bai yumi wokim wanpela pilai wantaim mani. Long dispela pilai yu ken kisim sampela mani na kisim i go long haus bilong yu na usim long laik bilong yu. Yu mas save, dispela mani em ino mani bilong Daavid, nogat. Em i manii bilong wanpela Uni (skul) is tap long Amerika na ol i bin givim bilong wokim dispela wok. Bihain bai dispela wok is tap long buk. I gat planti ol wokman bilong Uni ol i go aut long planti hap graun na long planti arapela kantri na mekim wankain wok. Bihain bai mipela mumutim olgeta wok na raitim wanpela buk. Translation: You and I will be 'playing' with money. When you play, you have the chance to earn money which you can take to your house and use as you like. You must know that this money does not belong to David (Tracer, primary investigator). It is money that belongs to a school (university) in the U.S. and it has been given to do this research. Later, (I will write about) this work in a book. There are many other workers from universities who are going out to various countries and doing the same kind of work. Later we'll put all the work we've done together and write a book. 101

PAGE 115

3. Narapela samting mi laik tokim yu-dispela em i wok tru tru, e mi no trik na em i no samting bilong paulim o giamanim yu. Translation: Another thing I'd like to talk to you about-this is real research (work), I'm not tricking you or pulling a joke on you. 4. Bipo yumi statim dispela wok, mi laik tokim yu bikpela samting. Taim yuk am long hia, yu no bin save wanem kain wok mi laik mekim. Sapos yu no gat laik long wokim, em orait, yu ken i go. Na sapos pilai i stat pinis na yu les long wokim, stile m i orait, yu ken i go. Translation: Before we start, I'd like to talk to you about something important. When you came here (today), you didn't know what kind ofwork I'd wanted to do (with you). If you don't like this work, it's alright, you can go. And if, after we start playing, you don't like the game, it's still alright, you can go. 5. Narapela samting: Bai wanwan maneri kam insait na wokim dispela pilai na bihain bai yu go aut na wet longpela taim liklik. Taim yu stap na wet yu ken toktok long olgeta samting tasollong dispela pilai yu bin mekim pinis yu no ken toktok. Sapos yu tokaut long wanem kain saamting yu bin mekim long pilai, yu ken bagarapim dispela wok. Translation: Another thing: one by one you (men and women) will come inside, play the game, and got outside and wait for awhile. While waiting, you can talk (to each other) about anything except about the game and what you did when you played. If you talk about this kind of thing (or), what you did in the game, you will really mess up this research. 6. Las samting e mi olsem: bilong helpim mi long wokim dispela wok bai olgeta wanwan manmeri kisim 2 kina. Em it ok tenkyu tasol em i no stap insait long pilai. Taim mi givim long yu, yu ken putim long sampela hap, e mi bilong yu nau. Translation: One last thing: for helping me with this research (work) I will give each of you 2 Kina. It's a just a thank you-it doesn't have to do with playing (the game). When I give it to you, you can put it somewhere (pocket/bag), it belongs to you now. 102

PAGE 116

1.2 Justice Game Script I. Long wokim dispela pilai, mi makim 3-pela manmeri, na bai 3-pela wok wantaim. [Olgeta 3-pela ol i bilong hia (dispela pies), tasol mi no inap kolim nem tru bilong ol na mi no inap tokaut husat i bin wok wantain husat narapela. Nogat. Na bihain tu, taim dispela wok i pinis, em bai stap olsem, mi no inap tokaut husat i bin wokim pilai wantaim husat narapela-na em bai stap olsem oltaim. Translation: To play the research (game), I will make groups of three (men and/or women), and the group of three will play together. The three people will all be from here (this place/village), but I won't use their real names and I won't tell anyone (playing) who the other players (members of the group) are (who played the game together). No. And also later, when the work is finished, I won't tell who played the game with any other person-and I'll keep it a secret always. 2. Orait, dispela 3-pela mi makim long wokim pilai wantaim, bai mi kolim wanpela "manmeri namba wan", narapela "manmeri namba tu", na narapela "manmeri namba tri". Na bai mi givim tenpela KI i go long manmeri namba wan na namba tu. Namba wan mas tingim pastaim na tokim mi olsem wanem e mi laik brukim o tilim dispela tenpela kina namellong en na namba tu. Namba wan em i ken salim KO na em yet holim olgeta KIO, o em salim KI i go long namba tuna em yet holim K9, o em salim K2 na em yet holim K8, K3 ... i go inap e mi salim KIO olgeta i go long namba tuna em yet bai kisim nogat olgeta. Orait, nau namba wan na namba tu ol no inap kisim kina yet. 01 i mas wet liklik inap namba tri em i pilai, na bihain bai mi tokim ol hamas kina ol i kisim. Translation: Alright, this trio that I make for the game, I will call one person "person number one" (PI), another "person number two" (P2), and the other person "person number three" (P3). And I will give ten KI (coins) to PI and P2. PI must first thing and then say how he/she would like to divide up the K 10 between himself and P2. PI can send KO and hold all K I 0, or he can send K I to P2 and hold K9, or send K2 and hold K8, K3 and hold K7, K4 ..... all the way until he can send all KIO to P2 and keep nothing at all. Alright, now PI and P2 are not finished yet (can't take the money yet). They must wait a little for P3 to play, and later I will tell them (PI and P2) how much Kina they will take. 3. Namba tri em bai pilai olsem: Bai mi givim faipela Kl long namba tri. Pastaim mi tokim em hamas namba wan i bin salim i go long long namba tu na hamas em yet holim na namba tri e mi ken wokim kainkain samting: ( 1) em i ken peim K 1 bilong 103

PAGE 117

rausim (tekwe) 3-pela kina bilong namba wan, (2) e mi ken peim Kl bilong skruim (addim) 3-pela kina i go long namba tu, (3) e mi ken peim K2 bilong rausim/tekwe 3-pela kina bilong namba wan na givim i go long namba tu, o (4) namba tri em i ken holim olgeta faipela kina bilong em yet na larim olgeta samting (kina bilong namba wan na namba tu) i stap wankain tasol. Translation: P2 will play like this: I will give five K1 (coins) (a total ofK5) to P3. First I will tell him/her how much P1 sent to P2 and how much P1 kept; and P3 can do these things: (1) he can pay K1 to take away K3 (that belong to) from P1, (2) he can pay K1 to add K3 to P2, (3) he can pay K2 to take away K3 from P1 and give them to P2, or (4) P3 can keep the five Kina that belong to him and leave it (the kina that belong to P 1 and P2 as decided by PI) just as it is. 104

PAGE 118

1.3 Standardized Examples (Used in the general assembly with pictorial representations.) 1.3.1 PI: Keep K3, Send K7 d) both Answers= a) PI: 3, P2: 7, P3: 5 b) PI: 0, P2: 7, P3: 4 c) PI: 3, P2: 10, P3: 4 d) Pl: 0, P2:10, P3: 3 1.3.2 PI: Keep K7, Send K3 d) both Answers= a)P1:7,P2:3,P3:5 b) PI: 4, P2: 3, P3: 4 c)Pl: 7,P2:6,P3:4 d) PI: 4, P2: 6, P3: 3 1.3.3 PI: Keep K5, Send K5 d) both Answers= a)P1:5,P2:5,P3:5 b) PI: 2, P2: 5, P3: 4 c) PI: 5, P2: 8, P3: 4 d) PI: 2, P2: 8, P3: 3 P3: a) nothing b) Kl to take away c) Kl to add P3: a) nothing b) K 1 to take away c) K 1 to add P3: a) nothing b) K l to take away c) Kl to add 105

PAGE 119

1.3.4 P1: Keep K10, Send KO d) both Answers= a) PI: 10,P2:0,P3:5 b) Pl: 7, P2: 0, P3: 4 c) Pl: 10, P2: 3, P3: 4 d) P1: 7, P2: 3, P3: 3 P3: a) nothing b) K1 to take away c) K1 to add 106

PAGE 120

1.4 Testing (Used after Player 1 entered the game enclosure to check understanding.) 1.4.1 PI: Keep K6, Send K4 both Answers= a)Pl:6,P2:4,P3: 5 b) PI: 3, P2: 4, P3: 4 c)P1:6,P2:7,P3:4 d) PI: 3, P2: 7, P3: 3 1.4.2 PI: Keep K4, Send K6 both Answers= a)P1:4,P2:6,P3: 5 b)Pl: l,P2:6,P3:4 c) PI: 4, P2: 9, P3: 4 .d) PI: 1, P2: 9, P3: 3 Extra if needed: 1.4.3 Pl: Keep K8, Send K2 both Answers= a)Pl:8,P2:2,P3: 5 b) PI: 5, P2: 2, P3: 4 c) PI: 8, P2: 5, P3: 4 d) PI: 5, P2: 5, P3: 3 P3: a) nothing b) K 1 to tekwe c) Kl to add d) P3: a) nothing b) Kl to tekwe c) Kl to add d) P3: a) nothing b) Kl to tekwe c) Kl to add d) 107

PAGE 121

1.5 Player Data Sheet ID Code ______ Village _____________ Sex. ____ Kolim nem bilong -------------------Kolim papa nem bilong y u? __ ___ Yu save wanem yia mama karim yu o hamas krismas bilong yu? (age) Yu marit o nogat? (married or not) (if yes) Wanpela meri tasol o moa? (#of wives) Hamas pikinnini bilong yu? (# of children) Yu skul o nogat? (attended school) (If yes) Yu pinisim wanem gred?(highest grade) Yu save go long lotu o nogat? (attend church) (if Y) Hamas taim long wanwan mun? (#times/mo.) Hamas gaden kaikai yu planim na gat nau? __ Yu save planim kopi? (you plant coffee) Kakao? (cocoa) Vanila? __ Yu save salim na kisim hamas kina long wanwan mun long k, k, o vanilla? ___ (from selling these cash crops, how much do you make a month?) Yu save wok bilong kisim kina sampela taim o stap long ples/wokim gaden tasol?_ (Do you work for wages some of the time or work you garden?) (If work) Bilong wok, yu save kisim hamas kina long wanwan mun? (how much do you earn per month?) *--------------------------------------------------------------------------------------* Role: Pt P2 P3 Associated Player IDs: Pt __________ P2 ________ P3 _______ ___ Pt Offer ______________________ P3 Action _________ Payouts: Pt __________ P2 __________ P3 _______ ___ Comments (level of understanding, etc): --------------108

PAGE 122

1.6 Post-Game Questions and Vignettes For Players 1: Long tingting bilong yu, hamas kina ol arapela manmeri i bin givirnlsalim i go long namba tu? (What do you think, how much [kina] did other people give/send to Player 2?) Bilong wanem? __ For Players 3: Long tingting bilong yu, ol arapela namba tri bin peim kina bilong rausim sampela kina bilong namba wan o skruim sampela kina long namba tu? (What do you think, did the other Players 3 pay kina to take away king from P1 or to add kina to P2?) Bilong wanem? __ Vignettes #1. Theft of Found Kina A man finds K 10 on the ground when walking through the bush. He takes it home, hides it, locks his house, and leaves. When he returns, someone has broken in and stolen it. a) the thief is jailed (follow up: what length jail term is fairest?) b) the thief is fined, and the fine goes to the state (follow up: what size fine is fairest?) c) the victim receives financial compensation from the state (follow up: what size compensation is fairest?) d) the victim receives financial compensation from the thief (follow up: what size compensation is fairest?) e) the thief is killed (follow up: by whom?) f) other (suggest ______ _J 109

PAGE 123

#2. Theft of Earned Kina A man hides K 10 that he earned in his house, locks the house, and leaves. When he returns, someone has broken in and stolen the K 10. a) the thief is jailed (follow up: what length jail term is fairest?) b) the thief is fined, and the fine goes to the state (follow up: what size fine is fairest?) c) the victim receives financial compensation from the state (follow up: what size compensation is fairest?) d) the victim receives financial compensation from the thief (follow up: what size compensation is fairest?) e) the thief is killed (follow up: by whom?) f) other (suggest _____ -' #3. Accidental Killing/Manslaughter While driving, a man hits another man with his car and kills him. a) the driver is jailed (follow up: what length jail term is fairest?) b) the driver is fined, and the fine goes to the state (follow up: what size fine is fairest?) c) the victim's clan receives financial compensation from the state (follow up: what size compensation is fairest?) d) the victim's clan receives financial compensation from the driver (follow up: what size compensation is fairest?) e) the driver is killed (follow up: by whom?) f) other (suggest _____ ___/ 110

PAGE 124

4. Homicide Two friends get into a fight. One becomes so angry he takes a gun and fatally shoots his friend. a) the shooter is jailed (follow up: what length jail term is fairest?) b) the shooter is fined, and the fine goes to the state (follow up: what size fine is fairest?) c) the victim's clan receives financial compensation from the state (follow up: what size compensation is fairest?) d) the victim's clan receives financial compensation from the shooter (follow up: what size compensation is fairest?) e) the shooter is killed (follow up: by whom?) f) other (suggest _____ _____/ I II

PAGE 125

APPENDIX C Human Subjects Review Board Approval University of Colorado at Denver and Health Sciences Center Human Subjects Research Committee lnstiMional Review Board Downtown Denver Campus Box 120, P.O. Box 173364 Denver, Colorado 80217-3364 Phone: 303-556-4060, Fax: 303-556-5855 DATE: February 18, 2005 TO: David Tracer FROM: Deborah Kellogg, HSRC Chair SUBJECT: Human Subjects Research Protocol #2005-083Prosociality and Justice: A Cross-Cultural Experimental Study Your protocol, with changes, has been approved as non-exempt and should pose no more than minimal risk. This approval is good for up to one year from this date. Your responsibilities as a researcher include: If you make changes to your research protocol or design you should contact the HSRC. You are responsible for maintaining all documentation of consent. Unless specified differently in your protocol, all data and consents should be maintained for three years. If you should encounter adverse human subjects issues, please contact us immediately. If your research continues beyond one year from the above date, contact the HSRC for an extension The HSRC may audit your documents at any time. Good Luck with your research. 112

PAGE 126

BIBLIOGRAPHY Alexander, J. McKenzie 2000 Evolutionary Explanations of Distributive Justice. Philosophy of Science 67:490-516. Alexander R.D. 1979 Darwinism and Human Affairs. Seattle: University of Washington Press. Austen-Smith, David and J.S. Banks 2002 Costly Signaling and Cheap Talk in Models of Political Influence. European Journal of Political Economy 18:263-280. Axelrod, R. 1984 The Evolution of Cooperation. New York: Basic Books, Inc. Axelrod Robert and William D. Hamilton 1981 The Evolution of Cooperation. Science 211:1 390-1396. Balasko, Yves 1988 Foundations of the Theory of General Equilibrium: Economic Theory, Econometrics, and Mathematical Economics. Orlando: Academic Press Inc. Banks, Cyndi 1998 Custom in the Courts: Criminal Law (Compensation) Act of Papua New Guinea. British Journal of Criminology 38(2):299-317. Barnett, Randy E. 1977 Restitution: A New Paradigm ofCriminal Justice. Ethics 87:279-301. Bartholdi, John J. III, C.A. Butler, and M.A. Trick 1986 More on the Evolution of Cooperation. The Journal of Conflict Resolution 30(1):129-140. Bennet, Christopher 2002 The Varieties of Retributive Experience. The Philosophical Quarterly 52(207): 145-163. 1 13

PAGE 127

Berg, Joyce, J. Dickhaut, and K. McCabe 1995 Trust, Reciprocity, and Social History. Games and Economic Behavior l 0:122-142. Bolton, Gary E., J. Brandts, E. Katok, A Ockenfels, and R. Zwick N.d. Testing Theories of Other-Regarding Behavior. Discussion papers on Strategic Interaction 2002-43, Max Planck Institute of Economics, Strategic Interaction Group. Bolton, Gary E., E. Katok, and R. Zwick 1998 Dictator Game Giving: Rules of Fairness Versus Acts of Kindness. International Journal of Game Theory 27:269-299. Bolton, Gary E., and A Ockenfels 2000 ERC: A Theory of Equity, Reciprocity, and Competition. American Economic Review 90(1):166-193. Bowles, Samuel, and H. Giotis 2002a Homo reciprocans. Nature 415:125-128. 2002b Social Capital and Community Governance. The Economic Journal 112:F419-F436. 2004 The Evolution of Strong Reciprocity: Cooperation in Heterogeneous Populations. Theoretical Population Biology 65:17-28. Boyd, Robert, H. Giotis, S. Bowles, and P.J. Richerson 2002 The Evolution of Altruistic Punishment. Proceedings of the National Academy of the Sciences 100(6):3531-3535. Boyd, Robert and P.J. Richerson 2005 The Origin and Evolution of Cultures. Oxford: University Press. Bradley, B.J. 1999 Levels of Selection, Altruism, and Primate Behavior. Quarterly Review of Biology 74(2):171-194. Braithwaite, John 1996 Restorative Justice and a Better Future. In A Restorative Justice Reader. G. Johnstone, ed. Pp.83-1 07. Portland: Willan Publishing. 114

PAGE 128

Brosnan, S., and F.B.M. de Waal 2002 A Proximate Perspective on Reciprocal Altruism. Human Nature 13( 1 ): 129152. Burling, Robbins 1962 Maximization Theories and the Study of Economic Anthropology. American Anthropologist 64( 4): 802-821. Camerer, Colin 1999 Behavioral Economics: Reunifying Psychology and Economics. Proceedings ofthe National Academy ofthe Sciences 96:10575-10577. 2003 Behavioral Game Theory: Experiments in Strategic Interaction. New York: Russell Sage Foundation. Cameron, L. 1995 Raising the Stakes in the Ultimatum Game: Experimental Evidence from Indonesia. Discussion Paper, Princeton University. Central Intelligence Agency (CIA) 2005 CIA World Factbook. Electronic document, www.odci.gov/cia/publications/factbook/geos/pp.html, accessed July 15. Chapais, B. 2001 Primate Nepotism: What is the Explanatory Value of Kin Selection? International Journal of Primatology 22(2):203-229. Chapais, B., L. Savard, and C. Gauthier 2001 Kin Selection and the Distribution of Altruism in Relation to Degree of Kinship in Japanese Macaques (Macacafuscata). Behavioral Ecology and Sociobiology 49:493-502. Charness, Gary, and M. Rabin 2002 Understanding Social Preferences with Simple Tests. The Quarterly Journal of Economics Aug:817 -869. Clark, Kenneth, and M. Sefton 2001 The Sequential Prisoner's Dilemma: Evidence on Reciprocation. The Economic Journal 111 :51-68. 115

PAGE 129

Cohen, Ronald L. 2001 Provocations of Restorative Justice. Social Justice Research 14(2):209-232. Conroy, G. 2005 The Emergence of Culture and the Origins ofthe Genus Homo. In Reconstructing Human Origins: A Modem Synthesis. 2nd edition. Pp 294-343. New York: WW Norton. Cook, Scott 1966 The Obselete "Anti-Market" Mentality: A Critique of the Substantive Approach to Economic Anthropology 68(2):323-345. Croson, Rachel, and N. Buchan 1999 Gender and Culture: International Experimental Evidence from Trust Games. Gender and Economic Transactions 89(2):386-391. Dalton, George 1969 Theoretical Issues in Economic Anthropology. Current Anthropology 1 0( 1 ): 63-102. Daly, Kathleen 2000 Revisiting the Relationship Between Retributive and Restorative Justice. In Restorative Justice: From Philosophy to Practice. H. Strang and J. Braithwaite, eds. Dartmouth: Aldershoot. Daly, M., and M. Wilson 1978 Sex, Evolution, and Behavior. 2nd edition. Belmont: Wadsworth Publishing Company. Darlington, P .J. Jr. 1972 Nonmathematical Models for Evolution of Altruism, and for Group Selection. Proceedings of the National Academy ofthe Sciences 69(2):293297. 1978 Altruism: Its Characteristics and Evolution. Proceedings of the National Academy ofthe Sciences 75(1):385-389. Darwin, Charles 1859 The Origin ofSpecies by Natural Selection. London: John Murray. 116

PAGE 130

Dawkins, R. 1976 The Selfish Gene. New York: Oxford University Press. De Waal, F. 1991. The Chimpanzee's Sense of Social Regularity and its Relation to the Human Sense of Justice. American Behavioral Scientist 34:335-349. Debreu, Gerard, and H. Scarf 1963 A Limit on the Core of an Economy. International Economic Review 4:235246. Diekmann, Andreas 2004 The Power of Reciprocity: Fairness, Reciprocity, and Stakes in Variants of the Dictator Game. Journal of Conflict Resolution 48( 4 ):487 -505. Eckel, Catherine C., and P.J. Grossman 1998 Are Women Less Selfish than Men?: Evidence from Dictator Experiments. The Economic Journa11 08(May):726-735. Falk, Armin, E. Fehr, and U. Fischbacher 2003 On the Nature of Fair Behavior. Economic Inquiry 41 (I ):20-26. Fehr, Ernst and U. Fischbacher 2002a The Nature of Human Altruism. Nature 425:785-791. 2002b Why Social Preferences Matter-the Impact of Non-selfish Motives on Competition, Cooperation, and Incentives. The Economic Journal 112:CI-C33. 2004a Social Norms and Human Cooperation. TRENDS in Cognitive Sciences 8( 4): 185-190. 2004b Third-party Punishment and Social Norms. Evolution and Human Behavior 25:63-87. Fehr, Ernst, U. Fischbacher, and S. Gachter 2002 Strong Reciprocity, Human Cooperation and the Enforcement of Social Norms. European Economic Review 42:845-859. 117

PAGE 131

Fehr, Ernst, U. Fischbacher, and E. Tougareva 2002 Do High Stakes and Competition Undermine Fairness? Evidence from Russia. Working Paper No. I20. Institute for Empirical Research in Economics, University of Zurich. Fehr, Ernst, and S. Giichter I 998 Reciprocity and Economics: the Economic Implications of Homo Reciprocans. European Economic Review 42:845-859. 2000a Cooperation and Punishment in Public Goods Experiments. The America Economic Review. 90(4):980-994. 2000b Fairness and Retaliation: the Economics of Reciprocity. Journal of Economic Perspectives I 4: I 59I 8 I. 2002 Altruistic Punishment in Humans. Nature 4 I 5: I 3 7I 40. Fehr, Ernst, and J. Henrich In press Is Strong Reciprocity a Maladaptation? In The Genetic and Cultural Evolution of Cooperation. P. Hammerstein, ed. Cambridge: MIT Press. Fehr, Ernst, and B. Rockenbach 2003 Detrimental Effects of Sanctions on Human Altruism. Nature 422(13):137I40. 2004 Human Altruism: Economic, Neural, and Evolutionary Perspectives. Current Opinion in Neurobiology I 4:784-790. Fehr, E., and K. Schmidt I 999 A Theory of Fairness, Competition, and Cooperation. The Quarterly Journal of Economics 8 I 7-868. Fisher, Arthur I 992 Sociobiology: Science or Ideology? Society 29(5):67-80. Fishman, Michael A. 2003 Indirect Reciprocity among Imperfect Individuals. Journal of Theoretical Biology 225:285-292. Fleagle, J.G. 1999 Primate Adaptation and Evolution. 2nd edition. San Diego: Academic Press. 118

PAGE 132

Foucault, Michel I977 Discipline and Punish: The Birth of the Prison. Alan Sheridan, trans. New York: Pantheon Books. Fowler, James H., T. Johnson, and 0. Smirnov 2004 Egalitarian Motive and Altruistic Punishment. Nature 433 :E I. Gintis, H., and S. Bowles 2003 The Origins of Human Cooperation. In The Genetic and Cultural Origins of Culture. P. Hammerstein, ed. Pp. I-17. Cambridge: MIT Press. Gintis, H., E.A. Smith, and S. Bowles 200 I Costly Signaling and Cooperation. Journal of Theoretical Biology 2I3: 103-II9. Gintis, Herbert 2000 Strong Reciprocity and Human Sociality. Journal of Theoretical Biology 206: I69-179. Gintis, Herbert, S. Bowles, R. Boyd, and E. Fehr 2003 Explaining Altruistic Behavior in Humans. Evolution and Human Behavior 24: I53-I72. Gouldner, Alvin W. I960 The Norm of Reciprocity: a Preliminary Statement. American Sociological Review 25: 16I-I78. Gowdy, John, and I. Seidl 2004 Economic Man and Selfish Genes: the Implications of Group Selection for Economic Valuation and Policy. Journal ofSocio-Economics 33:343-358. Griset, Pamela L. I991 Determinate Sentencing: the Promise and the Reality of Retributive Justice. Albany: State University ofNew York Press. Hamilton, W.D. 1964. The Genetical Theory of Social Behavior: I and II. Journal of Theoretical Biology 7:1-52. I972 Altruism and Related Phenomena, Mainly in Social Insects. Annual Review of Ecology and Systematics 3:193-232. I19

PAGE 133

1975 Innate Social Aptitudes of Man: an Approach from Evolutionary Genetics. In Biosocial Anthropology. R. Fox, ed. Pp. 133-153. London: Malaby Press. Hampton, Jean 1984 The Moral Education Theory of Punishment. Philosophy and Public Affairs 13:208-238. Henrich, J., R. Boyd, S. Bowles, C. Camerer, E. Fehr, H. Giotis, and R. McElreath 2001 In Search of Homo Economicus: Behavioral Experiments in 15 Small-scale Societies. The American Economic Review 91 (2):73-78. Henrich, Joseph. 2000 Does Culture Matter in Economic Behavior? Ultimatum Game Bargaining among the Machiguenga of the Peruvian Amazon. The American Economic Review 90(4):973-979. 2001 Cultural Group Selection, Coevolutionary Processes and Large-scale Cooperation. Journal of Economic Behavior & Organization 53:3-35. Henrich, Joseph, and R. Boyd 2001 Why People Punish Defectors: Weak Conformist Transmission Can Stabilize Costly Enforcement of Norms in Cooperative Dilemmas. Journal of Theoretical Biology 208:79-89. Henrich, Joseph, R. Boyd, S. Bowles, C. Camerer, E. Fehr, and H. Giotis, eds. 2004 Foundations of Human Sociality. Economic Experiments and Ethnographic Evidence from Fifteen Small-Scale Societies. Oxford: University Press. Hoffman, Elizabeth, K. McCabe, K. Shachat, and V. Smith 1994 Preferences, Property Rights, and Anonymity in Bargaining Games. Games and Economic Behavior 7:346-380. Hoffman, Elizabeth, K. McCabe, and V. Smith 1996 Social Distance and Other-regarding Behavior in Dictator Games. The American Economic Review 86(3):653-660. Holt, Charles A, and A.E. Roth 2004 The Nash Equilibrium: a Perspective. Proceedings of the National Academy ofthe Sciences 101(12):3999-4002. 120

PAGE 134

Humphrey, Nicholas 1997 Varieties of Altruism-and the Common Ground Between. Social Research 64(2): 199-210. Johnstone, Gerry, ed. 2003 A Restorative Justice Reader: Texts, Sources, Context. Portland: Willan Publishing. Kitcher, Philip 1993 The Evolution of Human Altruism. The Journal of Philosophy 90(10):497516. Kreps, David, and A. Rubinstein 1997 An Appreciation. In Classics in Game Theory. H.W. Kuhn, ed. Pp. xi-xv. Princeton: Princeton University Press. LeClair, Edward E. Jr. 1962 Economic Theory and Economic Anthropology. American Anthropologist 64(6):1179-1203. Lewis, Gilbert 1980 Day of Shining Red. An Essay on Understanding Ritual. Cambridge: Cambridge University Press. Macintyre, Ferren 2002 Was Religion a Kinship Surrogate? Journal of the America Academy of Religion 72(3):653-694. Marshall, Tony F. 1998 Restorative Justice: an Overview. In A Restorative Justice Reader. G. Johnstone, ed. Pp. 46-56. Portland: Willan Publishing. Matessi, Carlo, and S. Karlin 1984 On the Evolution of Altruism by Kin Selection. Proceedings ofthe National Academy of the Sciences 81:1754-1758. Maynard-Smith, J. 1982 Evolution and the Theory of Games. Cambridge: University Press. 121

PAGE 135

McAndrew, Francis T. 2002 New Evolutionary Perspectives on Altruism: Multilevel-Selection and Costly Signaling Theories. Current Directions in Psychological Science 11 (2):79-82. McGrew, W.C. 1998. Culture in Nonhuman Primates? Annual Review of Anthropology 27: 301328. Mohtashemi, Mojdeh, and L. Mui 2003 Evolution of Indirect Reciprocity by Social Information: the Role of Trust and Reputation in Evolution of Altruism. Journal of Theoretical Biology 223:523-531. Nash, John F. Jr. 1950a The Bargaining Problem. Econometrica 18:155-162. 1950b Equilibrium Points in N-person Games. Proceedings of the National Academy of the Sciences 36:48-49. New York Times 2005 From Kuwait: A Message of Hope. New York Times, September 14: A7. Nowak, Martin A., K.M. Page, and K. Sigmund 2000 Fairness Versus Reason in the Ultimatum Game. Science 289( 5485): 17731776. Nowak, Martin A., and K. Sigmund 1998 The Dynamics of Indirect Reciprocity. Journal of Theoretical Biology 194:561-574. Papua New Guinea National Statistics Office (PNGNSO) 2005 Census Data. Electronic document, http://www.nso.gov.pg/Pop _Soc_ %20Stats/popsoc.htm, accessed July 15. Polanyi, Karl, C.M. Arensberg, and H.W. Pearson, eds. 1957 Trade and Market in the Early Empires. Economies in History and Theory. Glencoe, Illinois: The Free Press. Quervain, Dominique J.-F., U. Fischbacher, V. Treyer, M. Schellhammer, U. Schnyder, A. Buck, and E. Fehr 2004 The Neural Basis of Altruistic Punishment. Science 305:1254-1258. 122

PAGE 136

Rabin, Matthew 1993 Incorporating Fairness into Game Theory and Economics. The American Economic Review 83(5):1281-1302. Rabinow, Paul, ed. 1984 The Foucault Reader. New York: Pantheon Books. Rappaport, Roy A. 1968 Pigs for the Ancestors. Ritual in the Ecology of a New Guinea People. New Haven: Yale University Press. Rawls, J. 1999 A Theory of Justice. Rev. ed. Cambridge: Harvard University Press. Reed, K.E., and L.R. Bidner 2004 Primate Communities: Past, Present and Possible Future. Yearbook of Physical Anthropology 47: 2-39. Riechmann, Thomas 2002 Relative Payoffs and Evolutionary Spite: Evolutionary Equilibriums in Games with Finitely Many Players. Discussion paper 260. University of Hannover. Rischer, Nicholas 2002 Fairness: Theory & Practice of Distributive Justice. New Brunswick: Transaction Publishers. Romp, Graham 1997 Game Theory. Introduction and Applications. New York: Oxford University Press, Inc. Roth, A.E, V. Prasnikar, M. Okuno-Fujiwara, and S. Zamir 1991 Bargaining and Market Behavior in Jerusalem, Ljubljana, Pittsburgh, and Tokyo: an Experimental Study. The American Economic Review 81(5):10681095. Sahlins, Marshall 1972 Stone Age Economics. New York: Aldine. 123

PAGE 137

Schroeder, David A, J.E. Steel, A.J. Woodell, and A.F. Bembenek 2003 Justice Within Social Dilemmas. Personality and Social Psychology Review 7(4):374-387. Seyfarth, R.M. and D.L. Cheney 1984 Grooming, Alliances, and Reciprocal Altruism in Vervet Monkeys. Nature 308:541-543. Shapley, L.S. 1953 A Value for n-person Games. In Classics in Game Theory. H. W. Kuhn, ed. Pp. 69-79. Princeton: Princeton University Press. Sigmund, Karl, E. Fehr, and M.A. Nowak 2002 The Economics of Fair Play. Scientific American, Jan.:82-87. Silk, J.B. 2004 The Evolution of Cooperation in Primate Groups. In Moral Sentiments and Material Interests: the Foundation of Cooperation in Economic Life. H. Gintis, S. Bowles, R. Boyd, and E. Fehr, eds. Cambridge: MIT Press. Sillitoe, Paul 1998 An Introduction to the Anthropology of Melanesia: Culture and Tradition. Cambridge: University Press. Smith, Adam 1868 [1863] An Inquiry into the Nature and Causes ofthe Wealth ofNations. Rev. edition. Edinburgh: Adam and Charles Black. Smith, M. 1976. Group Selection. Quarterly Review of Biology 51:277-283. Strier, K.B. 2000 Evolution and Social Behavior. In Primate Behavioral Ecology. 2nd edition. Pp. 94-134. Boston: Allyn and Bacon. Telser, L.G. 1995 The Ultimatum Game and the Law of Demand. The Economic Journal 1 05(433): 1519-1523. 124

PAGE 138

Terborgh, J. and C.H. Janson 1986 The Socioecology of Primate Groups. Annual Review of Ecology and Systematics 17:111-135. Tracer, David 1991 The Interaction ofNutrition and Fertility among Au Forager-Horticulturalists of Papua New Guinea. Ph.D. dissertation, Department of Anthropology, University of Michigan. 2003 Selfishness and Fairness in Economic and Evolutionary Perspective: an Experimental Economic Study in Papua New Guinea. Current Anthropology 44(3):432-438. 2004 Market Integration, Reciprocity, and Fairness in Rural Papua New Guinea: Results from a Two-village Ultimatum Game Study. In Foundations of Human Sociality: Economic Experiments and Ethnographic Evidence from Fifteen Small-Scale Societies. Henrich, J., R. Boyd, S. Bowles, C. Camerer, E. Fehr, and H. Gintis, eds. Pp. 232-259. Oxford: University Press. Trivers, R.L. 1971 The Evolution ofReciprocal Altruism. Quarterly Review of Biology 46:3557. Walgrave, Lode 2004 Has Restorative Justice Appropriately Responded to Retribution Theory and Impulses. In Critical Issues in Restorative Justice. H. Zehr and B. Toews, eds. Pp.47-60. New York: Criminal Justice Press. Walster, E., G.W. Walster, and E. Berscheid 1978 Equity: Theory and Research. Boston: Allyn & Bacon. Wedekind, Claus, and M. Milinski 1996 Human Cooperation in the Simultaneous and the Alternating Prisoner's Dilemma: Pavlov Versus Generous Tit-for-Tat. Proceedings ofthe National Academy of the Sciences 93:2686-2689. Wilson, David Sloan 1983 The Group Selection Controversy: History and Current Status. Annual Review ofEcological Systems 14:159-187. 125

PAGE 139

Zahavi, A. 1975 Mate Selection-A Selection for a Handicap. Journal of Theoretical Biology 53(1 ):205-14. 1977 The Cost of Honesty (further remarks on the handicap principle). Journal of Theoretical Biology 67(3):603-5. Zehr, Howard 1985 Retributive Justice, Restorative Justice. In A Restorative Justice Reader. G. Johnstone, ed. Pp. 69-82. Portland: Willan Publishing. ZimmerTamakoshi, Laura 1997 The Last Big Man: Development and Men's Discontents in the Papua New Guinea High lands. Oceania 68: 1 07-122. 126